Initial import

17 files changed, 7152 insertions, 0 deletions

diff --git a/drivers/net/ethernet/ibm/emac/Kconfig b/drivers/net/ethernet/ibm/emac/Kconfig
new file mode 100644
index 000000000..3f44a30e0
--- /dev/null
+++ b/drivers/net/ethernet/ibm/emac/Kconfig
@@ -0,0 +1,76 @@
+config IBM_EMAC
+	tristate "IBM EMAC Ethernet support"
+	depends on PPC_DCR
+	select CRC32
+	help
+	  This driver supports the IBM EMAC family of Ethernet controllers
+	  typically found on 4xx embedded PowerPC chips, but also on the
+	  Axon southbridge for Cell.
+
+config IBM_EMAC_RXB
+	int "Number of receive buffers"
+	depends on IBM_EMAC
+	default "128"
+
+config IBM_EMAC_TXB
+	int "Number of transmit buffers"
+	depends on IBM_EMAC
+	default "64"
+
+config IBM_EMAC_POLL_WEIGHT
+	int "MAL NAPI polling weight"
+	depends on IBM_EMAC
+	default "32"
+
+config IBM_EMAC_RX_COPY_THRESHOLD
+	int "RX skb copy threshold (bytes)"
+	depends on IBM_EMAC
+	default "256"
+
+config IBM_EMAC_RX_SKB_HEADROOM
+	int "Additional RX skb headroom (bytes)"
+	depends on IBM_EMAC
+	default "0"
+	help
+	  Additional receive skb headroom. Note, that driver
+	  will always reserve at least 2 bytes to make IP header
+	  aligned, so usually there is no need to add any additional
+	  headroom.
+
+	  If unsure, set to 0.
+
+config IBM_EMAC_DEBUG
+	bool "Debugging"
+	depends on IBM_EMAC
+	default n
+
+# The options below has to be select'ed by the respective
+# processor types or platforms
+
+config IBM_EMAC_ZMII
+	bool
+	default n
+
+config IBM_EMAC_RGMII
+	bool
+	default n
+
+config IBM_EMAC_TAH
+	bool
+	default n
+
+config IBM_EMAC_EMAC4
+	bool
+	default n
+
+config IBM_EMAC_NO_FLOW_CTRL
+	bool
+	default n
+
+config IBM_EMAC_MAL_CLR_ICINTSTAT
+	bool
+	default n
+
+config IBM_EMAC_MAL_COMMON_ERR
+	bool
+	default n
diff --git a/drivers/net/ethernet/ibm/emac/Makefile b/drivers/net/ethernet/ibm/emac/Makefile
new file mode 100644
index 000000000..eba21835d
--- /dev/null
+++ b/drivers/net/ethernet/ibm/emac/Makefile
@@ -0,0 +1,11 @@
+#
+# Makefile for the PowerPC 4xx on-chip ethernet driver
+#
+
+obj-$(CONFIG_IBM_EMAC) += ibm_emac.o
+
+ibm_emac-y := mal.o core.o phy.o
+ibm_emac-$(CONFIG_IBM_EMAC_ZMII) += zmii.o
+ibm_emac-$(CONFIG_IBM_EMAC_RGMII) += rgmii.o
+ibm_emac-$(CONFIG_IBM_EMAC_TAH) += tah.o
+ibm_emac-$(CONFIG_IBM_EMAC_DEBUG) += debug.o
diff --git a/drivers/net/ethernet/ibm/emac/core.c b/drivers/net/ethernet/ibm/emac/core.c
new file mode 100644
index 000000000..b9df0cbd0
--- /dev/null
+++ b/drivers/net/ethernet/ibm/emac/core.c
@@ -0,0 +1,3100 @@
+/*
+ * drivers/net/ethernet/ibm/emac/core.c
+ *
+ * Driver for PowerPC 4xx on-chip ethernet controller.
+ *
+ * Copyright 2007 Benjamin Herrenschmidt, IBM Corp.
+ *                <benh@kernel.crashing.org>
+ *
+ * Based on the arch/ppc version of the driver:
+ *
+ * Copyright (c) 2004, 2005 Zultys Technologies.
+ * Eugene Surovegin <eugene.surovegin@zultys.com> or <ebs@ebshome.net>
+ *
+ * Based on original work by
+ * 	Matt Porter <mporter@kernel.crashing.org>
+ *	(c) 2003 Benjamin Herrenschmidt <benh@kernel.crashing.org>
+ *      Armin Kuster <akuster@mvista.com>
+ * 	Johnnie Peters <jpeters@mvista.com>
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ *
+ */
+
+#include <linux/module.h>
+#include <linux/sched.h>
+#include <linux/string.h>
+#include <linux/errno.h>
+#include <linux/delay.h>
+#include <linux/types.h>
+#include <linux/pci.h>
+#include <linux/etherdevice.h>
+#include <linux/skbuff.h>
+#include <linux/crc32.h>
+#include <linux/ethtool.h>
+#include <linux/mii.h>
+#include <linux/bitops.h>
+#include <linux/workqueue.h>
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include <linux/of_irq.h>
+#include <linux/of_net.h>
+#include <linux/slab.h>
+
+#include <asm/processor.h>
+#include <asm/io.h>
+#include <asm/dma.h>
+#include <asm/uaccess.h>
+#include <asm/dcr.h>
+#include <asm/dcr-regs.h>
+
+#include "core.h"
+
+/*
+ * Lack of dma_unmap_???? calls is intentional.
+ *
+ * API-correct usage requires additional support state information to be
+ * maintained for every RX and TX buffer descriptor (BD). Unfortunately, due to
+ * EMAC design (e.g. TX buffer passed from network stack can be split into
+ * several BDs, dma_map_single/dma_map_page can be used to map particular BD),
+ * maintaining such information will add additional overhead.
+ * Current DMA API implementation for 4xx processors only ensures cache coherency
+ * and dma_unmap_???? routines are empty and are likely to stay this way.
+ * I decided to omit dma_unmap_??? calls because I don't want to add additional
+ * complexity just for the sake of following some abstract API, when it doesn't
+ * add any real benefit to the driver. I understand that this decision maybe
+ * controversial, but I really tried to make code API-correct and efficient
+ * at the same time and didn't come up with code I liked :(.                --ebs
+ */
+
+#define DRV_NAME        "emac"
+#define DRV_VERSION     "3.54"
+#define DRV_DESC        "PPC 4xx OCP EMAC driver"
+
+MODULE_DESCRIPTION(DRV_DESC);
+MODULE_AUTHOR
+    ("Eugene Surovegin <eugene.surovegin@zultys.com> or <ebs@ebshome.net>");
+MODULE_LICENSE("GPL");
+
+/* minimum number of free TX descriptors required to wake up TX process */
+#define EMAC_TX_WAKEUP_THRESH		(NUM_TX_BUFF / 4)
+
+/* If packet size is less than this number, we allocate small skb and copy packet
+ * contents into it instead of just sending original big skb up
+ */
+#define EMAC_RX_COPY_THRESH		CONFIG_IBM_EMAC_RX_COPY_THRESHOLD
+
+/* Since multiple EMACs share MDIO lines in various ways, we need
+ * to avoid re-using the same PHY ID in cases where the arch didn't
+ * setup precise phy_map entries
+ *
+ * XXX This is something that needs to be reworked as we can have multiple
+ * EMAC "sets" (multiple ASICs containing several EMACs) though we can
+ * probably require in that case to have explicit PHY IDs in the device-tree
+ */
+static u32 busy_phy_map;
+static DEFINE_MUTEX(emac_phy_map_lock);
+
+/* This is the wait queue used to wait on any event related to probe, that
+ * is discovery of MALs, other EMACs, ZMII/RGMIIs, etc...
+ */
+static DECLARE_WAIT_QUEUE_HEAD(emac_probe_wait);
+
+/* Having stable interface names is a doomed idea. However, it would be nice
+ * if we didn't have completely random interface names at boot too :-) It's
+ * just a matter of making everybody's life easier. Since we are doing
+ * threaded probing, it's a bit harder though. The base idea here is that
+ * we make up a list of all emacs in the device-tree before we register the
+ * driver. Every emac will then wait for the previous one in the list to
+ * initialize before itself. We should also keep that list ordered by
+ * cell_index.
+ * That list is only 4 entries long, meaning that additional EMACs don't
+ * get ordering guarantees unless EMAC_BOOT_LIST_SIZE is increased.
+ */
+
+#define EMAC_BOOT_LIST_SIZE	4
+static struct device_node *emac_boot_list[EMAC_BOOT_LIST_SIZE];
+
+/* How long should I wait for dependent devices ? */
+#define EMAC_PROBE_DEP_TIMEOUT	(HZ * 5)
+
+/* I don't want to litter system log with timeout errors
+ * when we have brain-damaged PHY.
+ */
+static inline void emac_report_timeout_error(struct emac_instance *dev,
+					     const char *error)
+{
+	if (emac_has_feature(dev, EMAC_FTR_440GX_PHY_CLK_FIX |
+				  EMAC_FTR_460EX_PHY_CLK_FIX |
+				  EMAC_FTR_440EP_PHY_CLK_FIX))
+		DBG(dev, "%s" NL, error);
+	else if (net_ratelimit())
+		printk(KERN_ERR "%s: %s\n", dev->ofdev->dev.of_node->full_name,
+			error);
+}
+
+/* EMAC PHY clock workaround:
+ * 440EP/440GR has more sane SDR0_MFR register implementation than 440GX,
+ * which allows controlling each EMAC clock
+ */
+static inline void emac_rx_clk_tx(struct emac_instance *dev)
+{
+#ifdef CONFIG_PPC_DCR_NATIVE
+	if (emac_has_feature(dev, EMAC_FTR_440EP_PHY_CLK_FIX))
+		dcri_clrset(SDR0, SDR0_MFR,
+			    0, SDR0_MFR_ECS >> dev->cell_index);
+#endif
+}
+
+static inline void emac_rx_clk_default(struct emac_instance *dev)
+{
+#ifdef CONFIG_PPC_DCR_NATIVE
+	if (emac_has_feature(dev, EMAC_FTR_440EP_PHY_CLK_FIX))
+		dcri_clrset(SDR0, SDR0_MFR,
+			    SDR0_MFR_ECS >> dev->cell_index, 0);
+#endif
+}
+
+/* PHY polling intervals */
+#define PHY_POLL_LINK_ON	HZ
+#define PHY_POLL_LINK_OFF	(HZ / 5)
+
+/* Graceful stop timeouts in us.
+ * We should allow up to 1 frame time (full-duplex, ignoring collisions)
+ */
+#define STOP_TIMEOUT_10		1230
+#define STOP_TIMEOUT_100	124
+#define STOP_TIMEOUT_1000	13
+#define STOP_TIMEOUT_1000_JUMBO	73
+
+static unsigned char default_mcast_addr[] = {
+	0x01, 0x80, 0xC2, 0x00, 0x00, 0x01
+};
+
+/* Please, keep in sync with struct ibm_emac_stats/ibm_emac_error_stats */
+static const char emac_stats_keys[EMAC_ETHTOOL_STATS_COUNT][ETH_GSTRING_LEN] = {
+	"rx_packets", "rx_bytes", "tx_packets", "tx_bytes", "rx_packets_csum",
+	"tx_packets_csum", "tx_undo", "rx_dropped_stack", "rx_dropped_oom",
+	"rx_dropped_error", "rx_dropped_resize", "rx_dropped_mtu",
+	"rx_stopped", "rx_bd_errors", "rx_bd_overrun", "rx_bd_bad_packet",
+	"rx_bd_runt_packet", "rx_bd_short_event", "rx_bd_alignment_error",
+	"rx_bd_bad_fcs", "rx_bd_packet_too_long", "rx_bd_out_of_range",
+	"rx_bd_in_range", "rx_parity", "rx_fifo_overrun", "rx_overrun",
+	"rx_bad_packet", "rx_runt_packet", "rx_short_event",
+	"rx_alignment_error", "rx_bad_fcs", "rx_packet_too_long",
+	"rx_out_of_range", "rx_in_range", "tx_dropped", "tx_bd_errors",
+	"tx_bd_bad_fcs", "tx_bd_carrier_loss", "tx_bd_excessive_deferral",
+	"tx_bd_excessive_collisions", "tx_bd_late_collision",
+	"tx_bd_multple_collisions", "tx_bd_single_collision",
+	"tx_bd_underrun", "tx_bd_sqe", "tx_parity", "tx_underrun", "tx_sqe",
+	"tx_errors"
+};
+
+static irqreturn_t emac_irq(int irq, void *dev_instance);
+static void emac_clean_tx_ring(struct emac_instance *dev);
+static void __emac_set_multicast_list(struct emac_instance *dev);
+
+static inline int emac_phy_supports_gige(int phy_mode)
+{
+	return  phy_mode == PHY_MODE_GMII ||
+		phy_mode == PHY_MODE_RGMII ||
+		phy_mode == PHY_MODE_SGMII ||
+		phy_mode == PHY_MODE_TBI ||
+		phy_mode == PHY_MODE_RTBI;
+}
+
+static inline int emac_phy_gpcs(int phy_mode)
+{
+	return  phy_mode == PHY_MODE_SGMII ||
+		phy_mode == PHY_MODE_TBI ||
+		phy_mode == PHY_MODE_RTBI;
+}
+
+static inline void emac_tx_enable(struct emac_instance *dev)
+{
+	struct emac_regs __iomem *p = dev->emacp;
+	u32 r;
+
+	DBG(dev, "tx_enable" NL);
+
+	r = in_be32(&p->mr0);
+	if (!(r & EMAC_MR0_TXE))
+		out_be32(&p->mr0, r | EMAC_MR0_TXE);
+}
+
+static void emac_tx_disable(struct emac_instance *dev)
+{
+	struct emac_regs __iomem *p = dev->emacp;
+	u32 r;
+
+	DBG(dev, "tx_disable" NL);
+
+	r = in_be32(&p->mr0);
+	if (r & EMAC_MR0_TXE) {
+		int n = dev->stop_timeout;
+		out_be32(&p->mr0, r & ~EMAC_MR0_TXE);
+		while (!(in_be32(&p->mr0) & EMAC_MR0_TXI) && n) {
+			udelay(1);
+			--n;
+		}
+		if (unlikely(!n))
+			emac_report_timeout_error(dev, "TX disable timeout");
+	}
+}
+
+static void emac_rx_enable(struct emac_instance *dev)
+{
+	struct emac_regs __iomem *p = dev->emacp;
+	u32 r;
+
+	if (unlikely(test_bit(MAL_COMMAC_RX_STOPPED, &dev->commac.flags)))
+		goto out;
+
+	DBG(dev, "rx_enable" NL);
+
+	r = in_be32(&p->mr0);
+	if (!(r & EMAC_MR0_RXE)) {
+		if (unlikely(!(r & EMAC_MR0_RXI))) {
+			/* Wait if previous async disable is still in progress */
+			int n = dev->stop_timeout;
+			while (!(r = in_be32(&p->mr0) & EMAC_MR0_RXI) && n) {
+				udelay(1);
+				--n;
+			}
+			if (unlikely(!n))
+				emac_report_timeout_error(dev,
+							  "RX disable timeout");
+		}
+		out_be32(&p->mr0, r | EMAC_MR0_RXE);
+	}
+ out:
+	;
+}
+
+static void emac_rx_disable(struct emac_instance *dev)
+{
+	struct emac_regs __iomem *p = dev->emacp;
+	u32 r;
+
+	DBG(dev, "rx_disable" NL);
+
+	r = in_be32(&p->mr0);
+	if (r & EMAC_MR0_RXE) {
+		int n = dev->stop_timeout;
+		out_be32(&p->mr0, r & ~EMAC_MR0_RXE);
+		while (!(in_be32(&p->mr0) & EMAC_MR0_RXI) && n) {
+			udelay(1);
+			--n;
+		}
+		if (unlikely(!n))
+			emac_report_timeout_error(dev, "RX disable timeout");
+	}
+}
+
+static inline void emac_netif_stop(struct emac_instance *dev)
+{
+	netif_tx_lock_bh(dev->ndev);
+	netif_addr_lock(dev->ndev);
+	dev->no_mcast = 1;
+	netif_addr_unlock(dev->ndev);
+	netif_tx_unlock_bh(dev->ndev);
+	dev->ndev->trans_start = jiffies;	/* prevent tx timeout */
+	mal_poll_disable(dev->mal, &dev->commac);
+	netif_tx_disable(dev->ndev);
+}
+
+static inline void emac_netif_start(struct emac_instance *dev)
+{
+	netif_tx_lock_bh(dev->ndev);
+	netif_addr_lock(dev->ndev);
+	dev->no_mcast = 0;
+	if (dev->mcast_pending && netif_running(dev->ndev))
+		__emac_set_multicast_list(dev);
+	netif_addr_unlock(dev->ndev);
+	netif_tx_unlock_bh(dev->ndev);
+
+	netif_wake_queue(dev->ndev);
+
+	/* NOTE: unconditional netif_wake_queue is only appropriate
+	 * so long as all callers are assured to have free tx slots
+	 * (taken from tg3... though the case where that is wrong is
+	 *  not terribly harmful)
+	 */
+	mal_poll_enable(dev->mal, &dev->commac);
+}
+
+static inline void emac_rx_disable_async(struct emac_instance *dev)
+{
+	struct emac_regs __iomem *p = dev->emacp;
+	u32 r;
+
+	DBG(dev, "rx_disable_async" NL);
+
+	r = in_be32(&p->mr0);
+	if (r & EMAC_MR0_RXE)
+		out_be32(&p->mr0, r & ~EMAC_MR0_RXE);
+}
+
+static int emac_reset(struct emac_instance *dev)
+{
+	struct emac_regs __iomem *p = dev->emacp;
+	int n = 20;
+
+	DBG(dev, "reset" NL);
+
+	if (!dev->reset_failed) {
+		/* 40x erratum suggests stopping RX channel before reset,
+		 * we stop TX as well
+		 */
+		emac_rx_disable(dev);
+		emac_tx_disable(dev);
+	}
+
+#ifdef CONFIG_PPC_DCR_NATIVE
+	/*
+	 * PPC460EX/GT Embedded Processor Advanced User's Manual
+	 * section 28.10.1 Mode Register 0 (EMACx_MR0) states:
+	 * Note: The PHY must provide a TX Clk in order to perform a soft reset
+	 * of the EMAC. If none is present, select the internal clock
+	 * (SDR0_ETH_CFG[EMACx_PHY_CLK] = 1).
+	 * After a soft reset, select the external clock.
+	 */
+	if (emac_has_feature(dev, EMAC_FTR_460EX_PHY_CLK_FIX)) {
+		if (dev->phy_address == 0xffffffff &&
+		    dev->phy_map == 0xffffffff) {
+			/* No PHY: select internal loop clock before reset */
+			dcri_clrset(SDR0, SDR0_ETH_CFG,
+				    0, SDR0_ETH_CFG_ECS << dev->cell_index);
+		} else {
+			/* PHY present: select external clock before reset */
+			dcri_clrset(SDR0, SDR0_ETH_CFG,
+				    SDR0_ETH_CFG_ECS << dev->cell_index, 0);
+		}
+	}
+#endif
+
+	out_be32(&p->mr0, EMAC_MR0_SRST);
+	while ((in_be32(&p->mr0) & EMAC_MR0_SRST) && n)
+		--n;
+
+#ifdef CONFIG_PPC_DCR_NATIVE
+	if (emac_has_feature(dev, EMAC_FTR_460EX_PHY_CLK_FIX)) {
+		if (dev->phy_address == 0xffffffff &&
+		    dev->phy_map == 0xffffffff) {
+			/* No PHY: restore external clock source after reset */
+			dcri_clrset(SDR0, SDR0_ETH_CFG,
+				    SDR0_ETH_CFG_ECS << dev->cell_index, 0);
+		}
+	}
+#endif
+
+	if (n) {
+		dev->reset_failed = 0;
+		return 0;
+	} else {
+		emac_report_timeout_error(dev, "reset timeout");
+		dev->reset_failed = 1;
+		return -ETIMEDOUT;
+	}
+}
+
+static void emac_hash_mc(struct emac_instance *dev)
+{
+	const int regs = EMAC_XAHT_REGS(dev);
+	u32 *gaht_base = emac_gaht_base(dev);
+	u32 gaht_temp[regs];
+	struct netdev_hw_addr *ha;
+	int i;
+
+	DBG(dev, "hash_mc %d" NL, netdev_mc_count(dev->ndev));
+
+	memset(gaht_temp, 0, sizeof (gaht_temp));
+
+	netdev_for_each_mc_addr(ha, dev->ndev) {
+		int slot, reg, mask;
+		DBG2(dev, "mc %pM" NL, ha->addr);
+
+		slot = EMAC_XAHT_CRC_TO_SLOT(dev,
+					     ether_crc(ETH_ALEN, ha->addr));
+		reg = EMAC_XAHT_SLOT_TO_REG(dev, slot);
+		mask = EMAC_XAHT_SLOT_TO_MASK(dev, slot);
+
+		gaht_temp[reg] |= mask;
+	}
+
+	for (i = 0; i < regs; i++)
+		out_be32(gaht_base + i, gaht_temp[i]);
+}
+
+static inline u32 emac_iff2rmr(struct net_device *ndev)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+	u32 r;
+
+	r = EMAC_RMR_SP | EMAC_RMR_SFCS | EMAC_RMR_IAE | EMAC_RMR_BAE;
+
+	if (emac_has_feature(dev, EMAC_FTR_EMAC4))
+	    r |= EMAC4_RMR_BASE;
+	else
+	    r |= EMAC_RMR_BASE;
+
+	if (ndev->flags & IFF_PROMISC)
+		r |= EMAC_RMR_PME;
+	else if (ndev->flags & IFF_ALLMULTI ||
+			 (netdev_mc_count(ndev) > EMAC_XAHT_SLOTS(dev)))
+		r |= EMAC_RMR_PMME;
+	else if (!netdev_mc_empty(ndev))
+		r |= EMAC_RMR_MAE;
+
+	if (emac_has_feature(dev, EMAC_APM821XX_REQ_JUMBO_FRAME_SIZE)) {
+		r &= ~EMAC4_RMR_MJS_MASK;
+		r |= EMAC4_RMR_MJS(ndev->mtu);
+	}
+
+	return r;
+}
+
+static u32 __emac_calc_base_mr1(struct emac_instance *dev, int tx_size, int rx_size)
+{
+	u32 ret = EMAC_MR1_VLE | EMAC_MR1_IST | EMAC_MR1_TR0_MULT;
+
+	DBG2(dev, "__emac_calc_base_mr1" NL);
+
+	switch(tx_size) {
+	case 2048:
+		ret |= EMAC_MR1_TFS_2K;
+		break;
+	default:
+		printk(KERN_WARNING "%s: Unknown Tx FIFO size %d\n",
+		       dev->ndev->name, tx_size);
+	}
+
+	switch(rx_size) {
+	case 16384:
+		ret |= EMAC_MR1_RFS_16K;
+		break;
+	case 4096:
+		ret |= EMAC_MR1_RFS_4K;
+		break;
+	default:
+		printk(KERN_WARNING "%s: Unknown Rx FIFO size %d\n",
+		       dev->ndev->name, rx_size);
+	}
+
+	return ret;
+}
+
+static u32 __emac4_calc_base_mr1(struct emac_instance *dev, int tx_size, int rx_size)
+{
+	u32 ret = EMAC_MR1_VLE | EMAC_MR1_IST | EMAC4_MR1_TR |
+		EMAC4_MR1_OBCI(dev->opb_bus_freq / 1000000);
+
+	DBG2(dev, "__emac4_calc_base_mr1" NL);
+
+	switch(tx_size) {
+	case 16384:
+		ret |= EMAC4_MR1_TFS_16K;
+		break;
+	case 4096:
+		ret |= EMAC4_MR1_TFS_4K;
+		break;
+	case 2048:
+		ret |= EMAC4_MR1_TFS_2K;
+		break;
+	default:
+		printk(KERN_WARNING "%s: Unknown Tx FIFO size %d\n",
+		       dev->ndev->name, tx_size);
+	}
+
+	switch(rx_size) {
+	case 16384:
+		ret |= EMAC4_MR1_RFS_16K;
+		break;
+	case 4096:
+		ret |= EMAC4_MR1_RFS_4K;
+		break;
+	case 2048:
+		ret |= EMAC4_MR1_RFS_2K;
+		break;
+	default:
+		printk(KERN_WARNING "%s: Unknown Rx FIFO size %d\n",
+		       dev->ndev->name, rx_size);
+	}
+
+	return ret;
+}
+
+static u32 emac_calc_base_mr1(struct emac_instance *dev, int tx_size, int rx_size)
+{
+	return emac_has_feature(dev, EMAC_FTR_EMAC4) ?
+		__emac4_calc_base_mr1(dev, tx_size, rx_size) :
+		__emac_calc_base_mr1(dev, tx_size, rx_size);
+}
+
+static inline u32 emac_calc_trtr(struct emac_instance *dev, unsigned int size)
+{
+	if (emac_has_feature(dev, EMAC_FTR_EMAC4))
+		return ((size >> 6) - 1) << EMAC_TRTR_SHIFT_EMAC4;
+	else
+		return ((size >> 6) - 1) << EMAC_TRTR_SHIFT;
+}
+
+static inline u32 emac_calc_rwmr(struct emac_instance *dev,
+				 unsigned int low, unsigned int high)
+{
+	if (emac_has_feature(dev, EMAC_FTR_EMAC4))
+		return (low << 22) | ( (high & 0x3ff) << 6);
+	else
+		return (low << 23) | ( (high & 0x1ff) << 7);
+}
+
+static int emac_configure(struct emac_instance *dev)
+{
+	struct emac_regs __iomem *p = dev->emacp;
+	struct net_device *ndev = dev->ndev;
+	int tx_size, rx_size, link = netif_carrier_ok(dev->ndev);
+	u32 r, mr1 = 0;
+
+	DBG(dev, "configure" NL);
+
+	if (!link) {
+		out_be32(&p->mr1, in_be32(&p->mr1)
+			 | EMAC_MR1_FDE | EMAC_MR1_ILE);
+		udelay(100);
+	} else if (emac_reset(dev) < 0)
+		return -ETIMEDOUT;
+
+	if (emac_has_feature(dev, EMAC_FTR_HAS_TAH))
+		tah_reset(dev->tah_dev);
+
+	DBG(dev, " link = %d duplex = %d, pause = %d, asym_pause = %d\n",
+	    link, dev->phy.duplex, dev->phy.pause, dev->phy.asym_pause);
+
+	/* Default fifo sizes */
+	tx_size = dev->tx_fifo_size;
+	rx_size = dev->rx_fifo_size;
+
+	/* No link, force loopback */
+	if (!link)
+		mr1 = EMAC_MR1_FDE | EMAC_MR1_ILE;
+
+	/* Check for full duplex */
+	else if (dev->phy.duplex == DUPLEX_FULL)
+		mr1 |= EMAC_MR1_FDE | EMAC_MR1_MWSW_001;
+
+	/* Adjust fifo sizes, mr1 and timeouts based on link speed */
+	dev->stop_timeout = STOP_TIMEOUT_10;
+	switch (dev->phy.speed) {
+	case SPEED_1000:
+		if (emac_phy_gpcs(dev->phy.mode)) {
+			mr1 |= EMAC_MR1_MF_1000GPCS | EMAC_MR1_MF_IPPA(
+				(dev->phy.gpcs_address != 0xffffffff) ?
+				 dev->phy.gpcs_address : dev->phy.address);
+
+			/* Put some arbitrary OUI, Manuf & Rev IDs so we can
+			 * identify this GPCS PHY later.
+			 */
+			out_be32(&p->u1.emac4.ipcr, 0xdeadbeef);
+		} else
+			mr1 |= EMAC_MR1_MF_1000;
+
+		/* Extended fifo sizes */
+		tx_size = dev->tx_fifo_size_gige;
+		rx_size = dev->rx_fifo_size_gige;
+
+		if (dev->ndev->mtu > ETH_DATA_LEN) {
+			if (emac_has_feature(dev, EMAC_FTR_EMAC4))
+				mr1 |= EMAC4_MR1_JPSM;
+			else
+				mr1 |= EMAC_MR1_JPSM;
+			dev->stop_timeout = STOP_TIMEOUT_1000_JUMBO;
+		} else
+			dev->stop_timeout = STOP_TIMEOUT_1000;
+		break;
+	case SPEED_100:
+		mr1 |= EMAC_MR1_MF_100;
+		dev->stop_timeout = STOP_TIMEOUT_100;
+		break;
+	default: /* make gcc happy */
+		break;
+	}
+
+	if (emac_has_feature(dev, EMAC_FTR_HAS_RGMII))
+		rgmii_set_speed(dev->rgmii_dev, dev->rgmii_port,
+				dev->phy.speed);
+	if (emac_has_feature(dev, EMAC_FTR_HAS_ZMII))
+		zmii_set_speed(dev->zmii_dev, dev->zmii_port, dev->phy.speed);
+
+	/* on 40x erratum forces us to NOT use integrated flow control,
+	 * let's hope it works on 44x ;)
+	 */
+	if (!emac_has_feature(dev, EMAC_FTR_NO_FLOW_CONTROL_40x) &&
+	    dev->phy.duplex == DUPLEX_FULL) {
+		if (dev->phy.pause)
+			mr1 |= EMAC_MR1_EIFC | EMAC_MR1_APP;
+		else if (dev->phy.asym_pause)
+			mr1 |= EMAC_MR1_APP;
+	}
+
+	/* Add base settings & fifo sizes & program MR1 */
+	mr1 |= emac_calc_base_mr1(dev, tx_size, rx_size);
+	out_be32(&p->mr1, mr1);
+
+	/* Set individual MAC address */
+	out_be32(&p->iahr, (ndev->dev_addr[0] << 8) | ndev->dev_addr[1]);
+	out_be32(&p->ialr, (ndev->dev_addr[2] << 24) |
+		 (ndev->dev_addr[3] << 16) | (ndev->dev_addr[4] << 8) |
+		 ndev->dev_addr[5]);
+
+	/* VLAN Tag Protocol ID */
+	out_be32(&p->vtpid, 0x8100);
+
+	/* Receive mode register */
+	r = emac_iff2rmr(ndev);
+	if (r & EMAC_RMR_MAE)
+		emac_hash_mc(dev);
+	out_be32(&p->rmr, r);
+
+	/* FIFOs thresholds */
+	if (emac_has_feature(dev, EMAC_FTR_EMAC4))
+		r = EMAC4_TMR1((dev->mal_burst_size / dev->fifo_entry_size) + 1,
+			       tx_size / 2 / dev->fifo_entry_size);
+	else
+		r = EMAC_TMR1((dev->mal_burst_size / dev->fifo_entry_size) + 1,
+			      tx_size / 2 / dev->fifo_entry_size);
+	out_be32(&p->tmr1, r);
+	out_be32(&p->trtr, emac_calc_trtr(dev, tx_size / 2));
+
+	/* PAUSE frame is sent when RX FIFO reaches its high-water mark,
+	   there should be still enough space in FIFO to allow the our link
+	   partner time to process this frame and also time to send PAUSE
+	   frame itself.
+
+	   Here is the worst case scenario for the RX FIFO "headroom"
+	   (from "The Switch Book") (100Mbps, without preamble, inter-frame gap):
+
+	   1) One maximum-length frame on TX                    1522 bytes
+	   2) One PAUSE frame time                                64 bytes
+	   3) PAUSE frame decode time allowance                   64 bytes
+	   4) One maximum-length frame on RX                    1522 bytes
+	   5) Round-trip propagation delay of the link (100Mb)    15 bytes
+	   ----------
+	   3187 bytes
+
+	   I chose to set high-water mark to RX_FIFO_SIZE / 4 (1024 bytes)
+	   low-water mark  to RX_FIFO_SIZE / 8 (512 bytes)
+	 */
+	r = emac_calc_rwmr(dev, rx_size / 8 / dev->fifo_entry_size,
+			   rx_size / 4 / dev->fifo_entry_size);
+	out_be32(&p->rwmr, r);
+
+	/* Set PAUSE timer to the maximum */
+	out_be32(&p->ptr, 0xffff);
+
+	/* IRQ sources */
+	r = EMAC_ISR_OVR | EMAC_ISR_BP | EMAC_ISR_SE |
+		EMAC_ISR_ALE | EMAC_ISR_BFCS | EMAC_ISR_PTLE | EMAC_ISR_ORE |
+		EMAC_ISR_IRE | EMAC_ISR_TE;
+	if (emac_has_feature(dev, EMAC_FTR_EMAC4))
+	    r |= EMAC4_ISR_TXPE | EMAC4_ISR_RXPE /* | EMAC4_ISR_TXUE |
+						  EMAC4_ISR_RXOE | */;
+	out_be32(&p->iser,  r);
+
+	/* We need to take GPCS PHY out of isolate mode after EMAC reset */
+	if (emac_phy_gpcs(dev->phy.mode)) {
+		if (dev->phy.gpcs_address != 0xffffffff)
+			emac_mii_reset_gpcs(&dev->phy);
+		else
+			emac_mii_reset_phy(&dev->phy);
+	}
+
+	return 0;
+}
+
+static void emac_reinitialize(struct emac_instance *dev)
+{
+	DBG(dev, "reinitialize" NL);
+
+	emac_netif_stop(dev);
+	if (!emac_configure(dev)) {
+		emac_tx_enable(dev);
+		emac_rx_enable(dev);
+	}
+	emac_netif_start(dev);
+}
+
+static void emac_full_tx_reset(struct emac_instance *dev)
+{
+	DBG(dev, "full_tx_reset" NL);
+
+	emac_tx_disable(dev);
+	mal_disable_tx_channel(dev->mal, dev->mal_tx_chan);
+	emac_clean_tx_ring(dev);
+	dev->tx_cnt = dev->tx_slot = dev->ack_slot = 0;
+
+	emac_configure(dev);
+
+	mal_enable_tx_channel(dev->mal, dev->mal_tx_chan);
+	emac_tx_enable(dev);
+	emac_rx_enable(dev);
+}
+
+static void emac_reset_work(struct work_struct *work)
+{
+	struct emac_instance *dev = container_of(work, struct emac_instance, reset_work);
+
+	DBG(dev, "reset_work" NL);
+
+	mutex_lock(&dev->link_lock);
+	if (dev->opened) {
+		emac_netif_stop(dev);
+		emac_full_tx_reset(dev);
+		emac_netif_start(dev);
+	}
+	mutex_unlock(&dev->link_lock);
+}
+
+static void emac_tx_timeout(struct net_device *ndev)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+
+	DBG(dev, "tx_timeout" NL);
+
+	schedule_work(&dev->reset_work);
+}
+
+
+static inline int emac_phy_done(struct emac_instance *dev, u32 stacr)
+{
+	int done = !!(stacr & EMAC_STACR_OC);
+
+	if (emac_has_feature(dev, EMAC_FTR_STACR_OC_INVERT))
+		done = !done;
+
+	return done;
+};
+
+static int __emac_mdio_read(struct emac_instance *dev, u8 id, u8 reg)
+{
+	struct emac_regs __iomem *p = dev->emacp;
+	u32 r = 0;
+	int n, err = -ETIMEDOUT;
+
+	mutex_lock(&dev->mdio_lock);
+
+	DBG2(dev, "mdio_read(%02x,%02x)" NL, id, reg);
+
+	/* Enable proper MDIO port */
+	if (emac_has_feature(dev, EMAC_FTR_HAS_ZMII))
+		zmii_get_mdio(dev->zmii_dev, dev->zmii_port);
+	if (emac_has_feature(dev, EMAC_FTR_HAS_RGMII))
+		rgmii_get_mdio(dev->rgmii_dev, dev->rgmii_port);
+
+	/* Wait for management interface to become idle */
+	n = 20;
+	while (!emac_phy_done(dev, in_be32(&p->stacr))) {
+		udelay(1);
+		if (!--n) {
+			DBG2(dev, " -> timeout wait idle\n");
+			goto bail;
+		}
+	}
+
+	/* Issue read command */
+	if (emac_has_feature(dev, EMAC_FTR_EMAC4))
+		r = EMAC4_STACR_BASE(dev->opb_bus_freq);
+	else
+		r = EMAC_STACR_BASE(dev->opb_bus_freq);
+	if (emac_has_feature(dev, EMAC_FTR_STACR_OC_INVERT))
+		r |= EMAC_STACR_OC;
+	if (emac_has_feature(dev, EMAC_FTR_HAS_NEW_STACR))
+		r |= EMACX_STACR_STAC_READ;
+	else
+		r |= EMAC_STACR_STAC_READ;
+	r |= (reg & EMAC_STACR_PRA_MASK)
+		| ((id & EMAC_STACR_PCDA_MASK) << EMAC_STACR_PCDA_SHIFT);
+	out_be32(&p->stacr, r);
+
+	/* Wait for read to complete */
+	n = 200;
+	while (!emac_phy_done(dev, (r = in_be32(&p->stacr)))) {
+		udelay(1);
+		if (!--n) {
+			DBG2(dev, " -> timeout wait complete\n");
+			goto bail;
+		}
+	}
+
+	if (unlikely(r & EMAC_STACR_PHYE)) {
+		DBG(dev, "mdio_read(%02x, %02x) failed" NL, id, reg);
+		err = -EREMOTEIO;
+		goto bail;
+	}
+
+	r = ((r >> EMAC_STACR_PHYD_SHIFT) & EMAC_STACR_PHYD_MASK);
+
+	DBG2(dev, "mdio_read -> %04x" NL, r);
+	err = 0;
+ bail:
+	if (emac_has_feature(dev, EMAC_FTR_HAS_RGMII))
+		rgmii_put_mdio(dev->rgmii_dev, dev->rgmii_port);
+	if (emac_has_feature(dev, EMAC_FTR_HAS_ZMII))
+		zmii_put_mdio(dev->zmii_dev, dev->zmii_port);
+	mutex_unlock(&dev->mdio_lock);
+
+	return err == 0 ? r : err;
+}
+
+static void __emac_mdio_write(struct emac_instance *dev, u8 id, u8 reg,
+			      u16 val)
+{
+	struct emac_regs __iomem *p = dev->emacp;
+	u32 r = 0;
+	int n, err = -ETIMEDOUT;
+
+	mutex_lock(&dev->mdio_lock);
+
+	DBG2(dev, "mdio_write(%02x,%02x,%04x)" NL, id, reg, val);
+
+	/* Enable proper MDIO port */
+	if (emac_has_feature(dev, EMAC_FTR_HAS_ZMII))
+		zmii_get_mdio(dev->zmii_dev, dev->zmii_port);
+	if (emac_has_feature(dev, EMAC_FTR_HAS_RGMII))
+		rgmii_get_mdio(dev->rgmii_dev, dev->rgmii_port);
+
+	/* Wait for management interface to be idle */
+	n = 20;
+	while (!emac_phy_done(dev, in_be32(&p->stacr))) {
+		udelay(1);
+		if (!--n) {
+			DBG2(dev, " -> timeout wait idle\n");
+			goto bail;
+		}
+	}
+
+	/* Issue write command */
+	if (emac_has_feature(dev, EMAC_FTR_EMAC4))
+		r = EMAC4_STACR_BASE(dev->opb_bus_freq);
+	else
+		r = EMAC_STACR_BASE(dev->opb_bus_freq);
+	if (emac_has_feature(dev, EMAC_FTR_STACR_OC_INVERT))
+		r |= EMAC_STACR_OC;
+	if (emac_has_feature(dev, EMAC_FTR_HAS_NEW_STACR))
+		r |= EMACX_STACR_STAC_WRITE;
+	else
+		r |= EMAC_STACR_STAC_WRITE;
+	r |= (reg & EMAC_STACR_PRA_MASK) |
+		((id & EMAC_STACR_PCDA_MASK) << EMAC_STACR_PCDA_SHIFT) |
+		(val << EMAC_STACR_PHYD_SHIFT);
+	out_be32(&p->stacr, r);
+
+	/* Wait for write to complete */
+	n = 200;
+	while (!emac_phy_done(dev, in_be32(&p->stacr))) {
+		udelay(1);
+		if (!--n) {
+			DBG2(dev, " -> timeout wait complete\n");
+			goto bail;
+		}
+	}
+	err = 0;
+ bail:
+	if (emac_has_feature(dev, EMAC_FTR_HAS_RGMII))
+		rgmii_put_mdio(dev->rgmii_dev, dev->rgmii_port);
+	if (emac_has_feature(dev, EMAC_FTR_HAS_ZMII))
+		zmii_put_mdio(dev->zmii_dev, dev->zmii_port);
+	mutex_unlock(&dev->mdio_lock);
+}
+
+static int emac_mdio_read(struct net_device *ndev, int id, int reg)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+	int res;
+
+	res = __emac_mdio_read((dev->mdio_instance &&
+				dev->phy.gpcs_address != id) ?
+				dev->mdio_instance : dev,
+			       (u8) id, (u8) reg);
+	return res;
+}
+
+static void emac_mdio_write(struct net_device *ndev, int id, int reg, int val)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+
+	__emac_mdio_write((dev->mdio_instance &&
+			   dev->phy.gpcs_address != id) ?
+			   dev->mdio_instance : dev,
+			  (u8) id, (u8) reg, (u16) val);
+}
+
+/* Tx lock BH */
+static void __emac_set_multicast_list(struct emac_instance *dev)
+{
+	struct emac_regs __iomem *p = dev->emacp;
+	u32 rmr = emac_iff2rmr(dev->ndev);
+
+	DBG(dev, "__multicast %08x" NL, rmr);
+
+	/* I decided to relax register access rules here to avoid
+	 * full EMAC reset.
+	 *
+	 * There is a real problem with EMAC4 core if we use MWSW_001 bit
+	 * in MR1 register and do a full EMAC reset.
+	 * One TX BD status update is delayed and, after EMAC reset, it
+	 * never happens, resulting in TX hung (it'll be recovered by TX
+	 * timeout handler eventually, but this is just gross).
+	 * So we either have to do full TX reset or try to cheat here :)
+	 *
+	 * The only required change is to RX mode register, so I *think* all
+	 * we need is just to stop RX channel. This seems to work on all
+	 * tested SoCs.                                                --ebs
+	 *
+	 * If we need the full reset, we might just trigger the workqueue
+	 * and do it async... a bit nasty but should work --BenH
+	 */
+	dev->mcast_pending = 0;
+	emac_rx_disable(dev);
+	if (rmr & EMAC_RMR_MAE)
+		emac_hash_mc(dev);
+	out_be32(&p->rmr, rmr);
+	emac_rx_enable(dev);
+}
+
+/* Tx lock BH */
+static void emac_set_multicast_list(struct net_device *ndev)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+
+	DBG(dev, "multicast" NL);
+
+	BUG_ON(!netif_running(dev->ndev));
+
+	if (dev->no_mcast) {
+		dev->mcast_pending = 1;
+		return;
+	}
+	__emac_set_multicast_list(dev);
+}
+
+static int emac_resize_rx_ring(struct emac_instance *dev, int new_mtu)
+{
+	int rx_sync_size = emac_rx_sync_size(new_mtu);
+	int rx_skb_size = emac_rx_skb_size(new_mtu);
+	int i, ret = 0;
+	int mr1_jumbo_bit_change = 0;
+
+	mutex_lock(&dev->link_lock);
+	emac_netif_stop(dev);
+	emac_rx_disable(dev);
+	mal_disable_rx_channel(dev->mal, dev->mal_rx_chan);
+
+	if (dev->rx_sg_skb) {
+		++dev->estats.rx_dropped_resize;
+		dev_kfree_skb(dev->rx_sg_skb);
+		dev->rx_sg_skb = NULL;
+	}
+
+	/* Make a first pass over RX ring and mark BDs ready, dropping
+	 * non-processed packets on the way. We need this as a separate pass
+	 * to simplify error recovery in the case of allocation failure later.
+	 */
+	for (i = 0; i < NUM_RX_BUFF; ++i) {
+		if (dev->rx_desc[i].ctrl & MAL_RX_CTRL_FIRST)
+			++dev->estats.rx_dropped_resize;
+
+		dev->rx_desc[i].data_len = 0;
+		dev->rx_desc[i].ctrl = MAL_RX_CTRL_EMPTY |
+		    (i == (NUM_RX_BUFF - 1) ? MAL_RX_CTRL_WRAP : 0);
+	}
+
+	/* Reallocate RX ring only if bigger skb buffers are required */
+	if (rx_skb_size <= dev->rx_skb_size)
+		goto skip;
+
+	/* Second pass, allocate new skbs */
+	for (i = 0; i < NUM_RX_BUFF; ++i) {
+		struct sk_buff *skb = alloc_skb(rx_skb_size, GFP_ATOMIC);
+		if (!skb) {
+			ret = -ENOMEM;
+			goto oom;
+		}
+
+		BUG_ON(!dev->rx_skb[i]);
+		dev_kfree_skb(dev->rx_skb[i]);
+
+		skb_reserve(skb, EMAC_RX_SKB_HEADROOM + 2);
+		dev->rx_desc[i].data_ptr =
+		    dma_map_single(&dev->ofdev->dev, skb->data - 2, rx_sync_size,
+				   DMA_FROM_DEVICE) + 2;
+		dev->rx_skb[i] = skb;
+	}
+ skip:
+	/* Check if we need to change "Jumbo" bit in MR1 */
+	if (emac_has_feature(dev, EMAC_APM821XX_REQ_JUMBO_FRAME_SIZE)) {
+		mr1_jumbo_bit_change = (new_mtu > ETH_DATA_LEN) ||
+				(dev->ndev->mtu > ETH_DATA_LEN);
+	} else {
+		mr1_jumbo_bit_change = (new_mtu > ETH_DATA_LEN) ^
+				(dev->ndev->mtu > ETH_DATA_LEN);
+	}
+
+	if (mr1_jumbo_bit_change) {
+		/* This is to prevent starting RX channel in emac_rx_enable() */
+		set_bit(MAL_COMMAC_RX_STOPPED, &dev->commac.flags);
+
+		dev->ndev->mtu = new_mtu;
+		emac_full_tx_reset(dev);
+	}
+
+	mal_set_rcbs(dev->mal, dev->mal_rx_chan, emac_rx_size(new_mtu));
+ oom:
+	/* Restart RX */
+	clear_bit(MAL_COMMAC_RX_STOPPED, &dev->commac.flags);
+	dev->rx_slot = 0;
+	mal_enable_rx_channel(dev->mal, dev->mal_rx_chan);
+	emac_rx_enable(dev);
+	emac_netif_start(dev);
+	mutex_unlock(&dev->link_lock);
+
+	return ret;
+}
+
+/* Process ctx, rtnl_lock semaphore */
+static int emac_change_mtu(struct net_device *ndev, int new_mtu)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+	int ret = 0;
+
+	if (new_mtu < EMAC_MIN_MTU || new_mtu > dev->max_mtu)
+		return -EINVAL;
+
+	DBG(dev, "change_mtu(%d)" NL, new_mtu);
+
+	if (netif_running(ndev)) {
+		/* Check if we really need to reinitialize RX ring */
+		if (emac_rx_skb_size(ndev->mtu) != emac_rx_skb_size(new_mtu))
+			ret = emac_resize_rx_ring(dev, new_mtu);
+	}
+
+	if (!ret) {
+		ndev->mtu = new_mtu;
+		dev->rx_skb_size = emac_rx_skb_size(new_mtu);
+		dev->rx_sync_size = emac_rx_sync_size(new_mtu);
+	}
+
+	return ret;
+}
+
+static void emac_clean_tx_ring(struct emac_instance *dev)
+{
+	int i;
+
+	for (i = 0; i < NUM_TX_BUFF; ++i) {
+		if (dev->tx_skb[i]) {
+			dev_kfree_skb(dev->tx_skb[i]);
+			dev->tx_skb[i] = NULL;
+			if (dev->tx_desc[i].ctrl & MAL_TX_CTRL_READY)
+				++dev->estats.tx_dropped;
+		}
+		dev->tx_desc[i].ctrl = 0;
+		dev->tx_desc[i].data_ptr = 0;
+	}
+}
+
+static void emac_clean_rx_ring(struct emac_instance *dev)
+{
+	int i;
+
+	for (i = 0; i < NUM_RX_BUFF; ++i)
+		if (dev->rx_skb[i]) {
+			dev->rx_desc[i].ctrl = 0;
+			dev_kfree_skb(dev->rx_skb[i]);
+			dev->rx_skb[i] = NULL;
+			dev->rx_desc[i].data_ptr = 0;
+		}
+
+	if (dev->rx_sg_skb) {
+		dev_kfree_skb(dev->rx_sg_skb);
+		dev->rx_sg_skb = NULL;
+	}
+}
+
+static inline int emac_alloc_rx_skb(struct emac_instance *dev, int slot,
+				    gfp_t flags)
+{
+	struct sk_buff *skb = alloc_skb(dev->rx_skb_size, flags);
+	if (unlikely(!skb))
+		return -ENOMEM;
+
+	dev->rx_skb[slot] = skb;
+	dev->rx_desc[slot].data_len = 0;
+
+	skb_reserve(skb, EMAC_RX_SKB_HEADROOM + 2);
+	dev->rx_desc[slot].data_ptr =
+	    dma_map_single(&dev->ofdev->dev, skb->data - 2, dev->rx_sync_size,
+			   DMA_FROM_DEVICE) + 2;
+	wmb();
+	dev->rx_desc[slot].ctrl = MAL_RX_CTRL_EMPTY |
+	    (slot == (NUM_RX_BUFF - 1) ? MAL_RX_CTRL_WRAP : 0);
+
+	return 0;
+}
+
+static void emac_print_link_status(struct emac_instance *dev)
+{
+	if (netif_carrier_ok(dev->ndev))
+		printk(KERN_INFO "%s: link is up, %d %s%s\n",
+		       dev->ndev->name, dev->phy.speed,
+		       dev->phy.duplex == DUPLEX_FULL ? "FDX" : "HDX",
+		       dev->phy.pause ? ", pause enabled" :
+		       dev->phy.asym_pause ? ", asymmetric pause enabled" : "");
+	else
+		printk(KERN_INFO "%s: link is down\n", dev->ndev->name);
+}
+
+/* Process ctx, rtnl_lock semaphore */
+static int emac_open(struct net_device *ndev)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+	int err, i;
+
+	DBG(dev, "open" NL);
+
+	/* Setup error IRQ handler */
+	err = request_irq(dev->emac_irq, emac_irq, 0, "EMAC", dev);
+	if (err) {
+		printk(KERN_ERR "%s: failed to request IRQ %d\n",
+		       ndev->name, dev->emac_irq);
+		return err;
+	}
+
+	/* Allocate RX ring */
+	for (i = 0; i < NUM_RX_BUFF; ++i)
+		if (emac_alloc_rx_skb(dev, i, GFP_KERNEL)) {
+			printk(KERN_ERR "%s: failed to allocate RX ring\n",
+			       ndev->name);
+			goto oom;
+		}
+
+	dev->tx_cnt = dev->tx_slot = dev->ack_slot = dev->rx_slot = 0;
+	clear_bit(MAL_COMMAC_RX_STOPPED, &dev->commac.flags);
+	dev->rx_sg_skb = NULL;
+
+	mutex_lock(&dev->link_lock);
+	dev->opened = 1;
+
+	/* Start PHY polling now.
+	 */
+	if (dev->phy.address >= 0) {
+		int link_poll_interval;
+		if (dev->phy.def->ops->poll_link(&dev->phy)) {
+			dev->phy.def->ops->read_link(&dev->phy);
+			emac_rx_clk_default(dev);
+			netif_carrier_on(dev->ndev);
+			link_poll_interval = PHY_POLL_LINK_ON;
+		} else {
+			emac_rx_clk_tx(dev);
+			netif_carrier_off(dev->ndev);
+			link_poll_interval = PHY_POLL_LINK_OFF;
+		}
+		dev->link_polling = 1;
+		wmb();
+		schedule_delayed_work(&dev->link_work, link_poll_interval);
+		emac_print_link_status(dev);
+	} else
+		netif_carrier_on(dev->ndev);
+
+	/* Required for Pause packet support in EMAC */
+	dev_mc_add_global(ndev, default_mcast_addr);
+
+	emac_configure(dev);
+	mal_poll_add(dev->mal, &dev->commac);
+	mal_enable_tx_channel(dev->mal, dev->mal_tx_chan);
+	mal_set_rcbs(dev->mal, dev->mal_rx_chan, emac_rx_size(ndev->mtu));
+	mal_enable_rx_channel(dev->mal, dev->mal_rx_chan);
+	emac_tx_enable(dev);
+	emac_rx_enable(dev);
+	emac_netif_start(dev);
+
+	mutex_unlock(&dev->link_lock);
+
+	return 0;
+ oom:
+	emac_clean_rx_ring(dev);
+	free_irq(dev->emac_irq, dev);
+
+	return -ENOMEM;
+}
+
+/* BHs disabled */
+#if 0
+static int emac_link_differs(struct emac_instance *dev)
+{
+	u32 r = in_be32(&dev->emacp->mr1);
+
+	int duplex = r & EMAC_MR1_FDE ? DUPLEX_FULL : DUPLEX_HALF;
+	int speed, pause, asym_pause;
+
+	if (r & EMAC_MR1_MF_1000)
+		speed = SPEED_1000;
+	else if (r & EMAC_MR1_MF_100)
+		speed = SPEED_100;
+	else
+		speed = SPEED_10;
+
+	switch (r & (EMAC_MR1_EIFC | EMAC_MR1_APP)) {
+	case (EMAC_MR1_EIFC | EMAC_MR1_APP):
+		pause = 1;
+		asym_pause = 0;
+		break;
+	case EMAC_MR1_APP:
+		pause = 0;
+		asym_pause = 1;
+		break;
+	default:
+		pause = asym_pause = 0;
+	}
+	return speed != dev->phy.speed || duplex != dev->phy.duplex ||
+	    pause != dev->phy.pause || asym_pause != dev->phy.asym_pause;
+}
+#endif
+
+static void emac_link_timer(struct work_struct *work)
+{
+	struct emac_instance *dev =
+		container_of(to_delayed_work(work),
+			     struct emac_instance, link_work);
+	int link_poll_interval;
+
+	mutex_lock(&dev->link_lock);
+	DBG2(dev, "link timer" NL);
+
+	if (!dev->opened)
+		goto bail;
+
+	if (dev->phy.def->ops->poll_link(&dev->phy)) {
+		if (!netif_carrier_ok(dev->ndev)) {
+			emac_rx_clk_default(dev);
+			/* Get new link parameters */
+			dev->phy.def->ops->read_link(&dev->phy);
+
+			netif_carrier_on(dev->ndev);
+			emac_netif_stop(dev);
+			emac_full_tx_reset(dev);
+			emac_netif_start(dev);
+			emac_print_link_status(dev);
+		}
+		link_poll_interval = PHY_POLL_LINK_ON;
+	} else {
+		if (netif_carrier_ok(dev->ndev)) {
+			emac_rx_clk_tx(dev);
+			netif_carrier_off(dev->ndev);
+			netif_tx_disable(dev->ndev);
+			emac_reinitialize(dev);
+			emac_print_link_status(dev);
+		}
+		link_poll_interval = PHY_POLL_LINK_OFF;
+	}
+	schedule_delayed_work(&dev->link_work, link_poll_interval);
+ bail:
+	mutex_unlock(&dev->link_lock);
+}
+
+static void emac_force_link_update(struct emac_instance *dev)
+{
+	netif_carrier_off(dev->ndev);
+	smp_rmb();
+	if (dev->link_polling) {
+		cancel_delayed_work_sync(&dev->link_work);
+		if (dev->link_polling)
+			schedule_delayed_work(&dev->link_work,  PHY_POLL_LINK_OFF);
+	}
+}
+
+/* Process ctx, rtnl_lock semaphore */
+static int emac_close(struct net_device *ndev)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+
+	DBG(dev, "close" NL);
+
+	if (dev->phy.address >= 0) {
+		dev->link_polling = 0;
+		cancel_delayed_work_sync(&dev->link_work);
+	}
+	mutex_lock(&dev->link_lock);
+	emac_netif_stop(dev);
+	dev->opened = 0;
+	mutex_unlock(&dev->link_lock);
+
+	emac_rx_disable(dev);
+	emac_tx_disable(dev);
+	mal_disable_rx_channel(dev->mal, dev->mal_rx_chan);
+	mal_disable_tx_channel(dev->mal, dev->mal_tx_chan);
+	mal_poll_del(dev->mal, &dev->commac);
+
+	emac_clean_tx_ring(dev);
+	emac_clean_rx_ring(dev);
+
+	free_irq(dev->emac_irq, dev);
+
+	netif_carrier_off(ndev);
+
+	return 0;
+}
+
+static inline u16 emac_tx_csum(struct emac_instance *dev,
+			       struct sk_buff *skb)
+{
+	if (emac_has_feature(dev, EMAC_FTR_HAS_TAH) &&
+		(skb->ip_summed == CHECKSUM_PARTIAL)) {
+		++dev->stats.tx_packets_csum;
+		return EMAC_TX_CTRL_TAH_CSUM;
+	}
+	return 0;
+}
+
+static inline int emac_xmit_finish(struct emac_instance *dev, int len)
+{
+	struct emac_regs __iomem *p = dev->emacp;
+	struct net_device *ndev = dev->ndev;
+
+	/* Send the packet out. If the if makes a significant perf
+	 * difference, then we can store the TMR0 value in "dev"
+	 * instead
+	 */
+	if (emac_has_feature(dev, EMAC_FTR_EMAC4))
+		out_be32(&p->tmr0, EMAC4_TMR0_XMIT);
+	else
+		out_be32(&p->tmr0, EMAC_TMR0_XMIT);
+
+	if (unlikely(++dev->tx_cnt == NUM_TX_BUFF)) {
+		netif_stop_queue(ndev);
+		DBG2(dev, "stopped TX queue" NL);
+	}
+
+	ndev->trans_start = jiffies;
+	++dev->stats.tx_packets;
+	dev->stats.tx_bytes += len;
+
+	return NETDEV_TX_OK;
+}
+
+/* Tx lock BH */
+static int emac_start_xmit(struct sk_buff *skb, struct net_device *ndev)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+	unsigned int len = skb->len;
+	int slot;
+
+	u16 ctrl = EMAC_TX_CTRL_GFCS | EMAC_TX_CTRL_GP | MAL_TX_CTRL_READY |
+	    MAL_TX_CTRL_LAST | emac_tx_csum(dev, skb);
+
+	slot = dev->tx_slot++;
+	if (dev->tx_slot == NUM_TX_BUFF) {
+		dev->tx_slot = 0;
+		ctrl |= MAL_TX_CTRL_WRAP;
+	}
+
+	DBG2(dev, "xmit(%u) %d" NL, len, slot);
+
+	dev->tx_skb[slot] = skb;
+	dev->tx_desc[slot].data_ptr = dma_map_single(&dev->ofdev->dev,
+						     skb->data, len,
+						     DMA_TO_DEVICE);
+	dev->tx_desc[slot].data_len = (u16) len;
+	wmb();
+	dev->tx_desc[slot].ctrl = ctrl;
+
+	return emac_xmit_finish(dev, len);
+}
+
+static inline int emac_xmit_split(struct emac_instance *dev, int slot,
+				  u32 pd, int len, int last, u16 base_ctrl)
+{
+	while (1) {
+		u16 ctrl = base_ctrl;
+		int chunk = min(len, MAL_MAX_TX_SIZE);
+		len -= chunk;
+
+		slot = (slot + 1) % NUM_TX_BUFF;
+
+		if (last && !len)
+			ctrl |= MAL_TX_CTRL_LAST;
+		if (slot == NUM_TX_BUFF - 1)
+			ctrl |= MAL_TX_CTRL_WRAP;
+
+		dev->tx_skb[slot] = NULL;
+		dev->tx_desc[slot].data_ptr = pd;
+		dev->tx_desc[slot].data_len = (u16) chunk;
+		dev->tx_desc[slot].ctrl = ctrl;
+		++dev->tx_cnt;
+
+		if (!len)
+			break;
+
+		pd += chunk;
+	}
+	return slot;
+}
+
+/* Tx lock BH disabled (SG version for TAH equipped EMACs) */
+static int emac_start_xmit_sg(struct sk_buff *skb, struct net_device *ndev)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+	int nr_frags = skb_shinfo(skb)->nr_frags;
+	int len = skb->len, chunk;
+	int slot, i;
+	u16 ctrl;
+	u32 pd;
+
+	/* This is common "fast" path */
+	if (likely(!nr_frags && len <= MAL_MAX_TX_SIZE))
+		return emac_start_xmit(skb, ndev);
+
+	len -= skb->data_len;
+
+	/* Note, this is only an *estimation*, we can still run out of empty
+	 * slots because of the additional fragmentation into
+	 * MAL_MAX_TX_SIZE-sized chunks
+	 */
+	if (unlikely(dev->tx_cnt + nr_frags + mal_tx_chunks(len) > NUM_TX_BUFF))
+		goto stop_queue;
+
+	ctrl = EMAC_TX_CTRL_GFCS | EMAC_TX_CTRL_GP | MAL_TX_CTRL_READY |
+	    emac_tx_csum(dev, skb);
+	slot = dev->tx_slot;
+
+	/* skb data */
+	dev->tx_skb[slot] = NULL;
+	chunk = min(len, MAL_MAX_TX_SIZE);
+	dev->tx_desc[slot].data_ptr = pd =
+	    dma_map_single(&dev->ofdev->dev, skb->data, len, DMA_TO_DEVICE);
+	dev->tx_desc[slot].data_len = (u16) chunk;
+	len -= chunk;
+	if (unlikely(len))
+		slot = emac_xmit_split(dev, slot, pd + chunk, len, !nr_frags,
+				       ctrl);
+	/* skb fragments */
+	for (i = 0; i < nr_frags; ++i) {
+		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i];
+		len = skb_frag_size(frag);
+
+		if (unlikely(dev->tx_cnt + mal_tx_chunks(len) >= NUM_TX_BUFF))
+			goto undo_frame;
+
+		pd = skb_frag_dma_map(&dev->ofdev->dev, frag, 0, len,
+				      DMA_TO_DEVICE);
+
+		slot = emac_xmit_split(dev, slot, pd, len, i == nr_frags - 1,
+				       ctrl);
+	}
+
+	DBG2(dev, "xmit_sg(%u) %d - %d" NL, skb->len, dev->tx_slot, slot);
+
+	/* Attach skb to the last slot so we don't release it too early */
+	dev->tx_skb[slot] = skb;
+
+	/* Send the packet out */
+	if (dev->tx_slot == NUM_TX_BUFF - 1)
+		ctrl |= MAL_TX_CTRL_WRAP;
+	wmb();
+	dev->tx_desc[dev->tx_slot].ctrl = ctrl;
+	dev->tx_slot = (slot + 1) % NUM_TX_BUFF;
+
+	return emac_xmit_finish(dev, skb->len);
+
+ undo_frame:
+	/* Well, too bad. Our previous estimation was overly optimistic.
+	 * Undo everything.
+	 */
+	while (slot != dev->tx_slot) {
+		dev->tx_desc[slot].ctrl = 0;
+		--dev->tx_cnt;
+		if (--slot < 0)
+			slot = NUM_TX_BUFF - 1;
+	}
+	++dev->estats.tx_undo;
+
+ stop_queue:
+	netif_stop_queue(ndev);
+	DBG2(dev, "stopped TX queue" NL);
+	return NETDEV_TX_BUSY;
+}
+
+/* Tx lock BHs */
+static void emac_parse_tx_error(struct emac_instance *dev, u16 ctrl)
+{
+	struct emac_error_stats *st = &dev->estats;
+
+	DBG(dev, "BD TX error %04x" NL, ctrl);
+
+	++st->tx_bd_errors;
+	if (ctrl & EMAC_TX_ST_BFCS)
+		++st->tx_bd_bad_fcs;
+	if (ctrl & EMAC_TX_ST_LCS)
+		++st->tx_bd_carrier_loss;
+	if (ctrl & EMAC_TX_ST_ED)
+		++st->tx_bd_excessive_deferral;
+	if (ctrl & EMAC_TX_ST_EC)
+		++st->tx_bd_excessive_collisions;
+	if (ctrl & EMAC_TX_ST_LC)
+		++st->tx_bd_late_collision;
+	if (ctrl & EMAC_TX_ST_MC)
+		++st->tx_bd_multple_collisions;
+	if (ctrl & EMAC_TX_ST_SC)
+		++st->tx_bd_single_collision;
+	if (ctrl & EMAC_TX_ST_UR)
+		++st->tx_bd_underrun;
+	if (ctrl & EMAC_TX_ST_SQE)
+		++st->tx_bd_sqe;
+}
+
+static void emac_poll_tx(void *param)
+{
+	struct emac_instance *dev = param;
+	u32 bad_mask;
+
+	DBG2(dev, "poll_tx, %d %d" NL, dev->tx_cnt, dev->ack_slot);
+
+	if (emac_has_feature(dev, EMAC_FTR_HAS_TAH))
+		bad_mask = EMAC_IS_BAD_TX_TAH;
+	else
+		bad_mask = EMAC_IS_BAD_TX;
+
+	netif_tx_lock_bh(dev->ndev);
+	if (dev->tx_cnt) {
+		u16 ctrl;
+		int slot = dev->ack_slot, n = 0;
+	again:
+		ctrl = dev->tx_desc[slot].ctrl;
+		if (!(ctrl & MAL_TX_CTRL_READY)) {
+			struct sk_buff *skb = dev->tx_skb[slot];
+			++n;
+
+			if (skb) {
+				dev_kfree_skb(skb);
+				dev->tx_skb[slot] = NULL;
+			}
+			slot = (slot + 1) % NUM_TX_BUFF;
+
+			if (unlikely(ctrl & bad_mask))
+				emac_parse_tx_error(dev, ctrl);
+
+			if (--dev->tx_cnt)
+				goto again;
+		}
+		if (n) {
+			dev->ack_slot = slot;
+			if (netif_queue_stopped(dev->ndev) &&
+			    dev->tx_cnt < EMAC_TX_WAKEUP_THRESH)
+				netif_wake_queue(dev->ndev);
+
+			DBG2(dev, "tx %d pkts" NL, n);
+		}
+	}
+	netif_tx_unlock_bh(dev->ndev);
+}
+
+static inline void emac_recycle_rx_skb(struct emac_instance *dev, int slot,
+				       int len)
+{
+	struct sk_buff *skb = dev->rx_skb[slot];
+
+	DBG2(dev, "recycle %d %d" NL, slot, len);
+
+	if (len)
+		dma_map_single(&dev->ofdev->dev, skb->data - 2,
+			       EMAC_DMA_ALIGN(len + 2), DMA_FROM_DEVICE);
+
+	dev->rx_desc[slot].data_len = 0;
+	wmb();
+	dev->rx_desc[slot].ctrl = MAL_RX_CTRL_EMPTY |
+	    (slot == (NUM_RX_BUFF - 1) ? MAL_RX_CTRL_WRAP : 0);
+}
+
+static void emac_parse_rx_error(struct emac_instance *dev, u16 ctrl)
+{
+	struct emac_error_stats *st = &dev->estats;
+
+	DBG(dev, "BD RX error %04x" NL, ctrl);
+
+	++st->rx_bd_errors;
+	if (ctrl & EMAC_RX_ST_OE)
+		++st->rx_bd_overrun;
+	if (ctrl & EMAC_RX_ST_BP)
+		++st->rx_bd_bad_packet;
+	if (ctrl & EMAC_RX_ST_RP)
+		++st->rx_bd_runt_packet;
+	if (ctrl & EMAC_RX_ST_SE)
+		++st->rx_bd_short_event;
+	if (ctrl & EMAC_RX_ST_AE)
+		++st->rx_bd_alignment_error;
+	if (ctrl & EMAC_RX_ST_BFCS)
+		++st->rx_bd_bad_fcs;
+	if (ctrl & EMAC_RX_ST_PTL)
+		++st->rx_bd_packet_too_long;
+	if (ctrl & EMAC_RX_ST_ORE)
+		++st->rx_bd_out_of_range;
+	if (ctrl & EMAC_RX_ST_IRE)
+		++st->rx_bd_in_range;
+}
+
+static inline void emac_rx_csum(struct emac_instance *dev,
+				struct sk_buff *skb, u16 ctrl)
+{
+#ifdef CONFIG_IBM_EMAC_TAH
+	if (!ctrl && dev->tah_dev) {
+		skb->ip_summed = CHECKSUM_UNNECESSARY;
+		++dev->stats.rx_packets_csum;
+	}
+#endif
+}
+
+static inline int emac_rx_sg_append(struct emac_instance *dev, int slot)
+{
+	if (likely(dev->rx_sg_skb != NULL)) {
+		int len = dev->rx_desc[slot].data_len;
+		int tot_len = dev->rx_sg_skb->len + len;
+
+		if (unlikely(tot_len + 2 > dev->rx_skb_size)) {
+			++dev->estats.rx_dropped_mtu;
+			dev_kfree_skb(dev->rx_sg_skb);
+			dev->rx_sg_skb = NULL;
+		} else {
+			memcpy(skb_tail_pointer(dev->rx_sg_skb),
+					 dev->rx_skb[slot]->data, len);
+			skb_put(dev->rx_sg_skb, len);
+			emac_recycle_rx_skb(dev, slot, len);
+			return 0;
+		}
+	}
+	emac_recycle_rx_skb(dev, slot, 0);
+	return -1;
+}
+
+/* NAPI poll context */
+static int emac_poll_rx(void *param, int budget)
+{
+	struct emac_instance *dev = param;
+	int slot = dev->rx_slot, received = 0;
+
+	DBG2(dev, "poll_rx(%d)" NL, budget);
+
+ again:
+	while (budget > 0) {
+		int len;
+		struct sk_buff *skb;
+		u16 ctrl = dev->rx_desc[slot].ctrl;
+
+		if (ctrl & MAL_RX_CTRL_EMPTY)
+			break;
+
+		skb = dev->rx_skb[slot];
+		mb();
+		len = dev->rx_desc[slot].data_len;
+
+		if (unlikely(!MAL_IS_SINGLE_RX(ctrl)))
+			goto sg;
+
+		ctrl &= EMAC_BAD_RX_MASK;
+		if (unlikely(ctrl && ctrl != EMAC_RX_TAH_BAD_CSUM)) {
+			emac_parse_rx_error(dev, ctrl);
+			++dev->estats.rx_dropped_error;
+			emac_recycle_rx_skb(dev, slot, 0);
+			len = 0;
+			goto next;
+		}
+
+		if (len < ETH_HLEN) {
+			++dev->estats.rx_dropped_stack;
+			emac_recycle_rx_skb(dev, slot, len);
+			goto next;
+		}
+
+		if (len && len < EMAC_RX_COPY_THRESH) {
+			struct sk_buff *copy_skb =
+			    alloc_skb(len + EMAC_RX_SKB_HEADROOM + 2, GFP_ATOMIC);
+			if (unlikely(!copy_skb))
+				goto oom;
+
+			skb_reserve(copy_skb, EMAC_RX_SKB_HEADROOM + 2);
+			memcpy(copy_skb->data - 2, skb->data - 2, len + 2);
+			emac_recycle_rx_skb(dev, slot, len);
+			skb = copy_skb;
+		} else if (unlikely(emac_alloc_rx_skb(dev, slot, GFP_ATOMIC)))
+			goto oom;
+
+		skb_put(skb, len);
+	push_packet:
+		skb->protocol = eth_type_trans(skb, dev->ndev);
+		emac_rx_csum(dev, skb, ctrl);
+
+		if (unlikely(netif_receive_skb(skb) == NET_RX_DROP))
+			++dev->estats.rx_dropped_stack;
+	next:
+		++dev->stats.rx_packets;
+	skip:
+		dev->stats.rx_bytes += len;
+		slot = (slot + 1) % NUM_RX_BUFF;
+		--budget;
+		++received;
+		continue;
+	sg:
+		if (ctrl & MAL_RX_CTRL_FIRST) {
+			BUG_ON(dev->rx_sg_skb);
+			if (unlikely(emac_alloc_rx_skb(dev, slot, GFP_ATOMIC))) {
+				DBG(dev, "rx OOM %d" NL, slot);
+				++dev->estats.rx_dropped_oom;
+				emac_recycle_rx_skb(dev, slot, 0);
+			} else {
+				dev->rx_sg_skb = skb;
+				skb_put(skb, len);
+			}
+		} else if (!emac_rx_sg_append(dev, slot) &&
+			   (ctrl & MAL_RX_CTRL_LAST)) {
+
+			skb = dev->rx_sg_skb;
+			dev->rx_sg_skb = NULL;
+
+			ctrl &= EMAC_BAD_RX_MASK;
+			if (unlikely(ctrl && ctrl != EMAC_RX_TAH_BAD_CSUM)) {
+				emac_parse_rx_error(dev, ctrl);
+				++dev->estats.rx_dropped_error;
+				dev_kfree_skb(skb);
+				len = 0;
+			} else
+				goto push_packet;
+		}
+		goto skip;
+	oom:
+		DBG(dev, "rx OOM %d" NL, slot);
+		/* Drop the packet and recycle skb */
+		++dev->estats.rx_dropped_oom;
+		emac_recycle_rx_skb(dev, slot, 0);
+		goto next;
+	}
+
+	if (received) {
+		DBG2(dev, "rx %d BDs" NL, received);
+		dev->rx_slot = slot;
+	}
+
+	if (unlikely(budget && test_bit(MAL_COMMAC_RX_STOPPED, &dev->commac.flags))) {
+		mb();
+		if (!(dev->rx_desc[slot].ctrl & MAL_RX_CTRL_EMPTY)) {
+			DBG2(dev, "rx restart" NL);
+			received = 0;
+			goto again;
+		}
+
+		if (dev->rx_sg_skb) {
+			DBG2(dev, "dropping partial rx packet" NL);
+			++dev->estats.rx_dropped_error;
+			dev_kfree_skb(dev->rx_sg_skb);
+			dev->rx_sg_skb = NULL;
+		}
+
+		clear_bit(MAL_COMMAC_RX_STOPPED, &dev->commac.flags);
+		mal_enable_rx_channel(dev->mal, dev->mal_rx_chan);
+		emac_rx_enable(dev);
+		dev->rx_slot = 0;
+	}
+	return received;
+}
+
+/* NAPI poll context */
+static int emac_peek_rx(void *param)
+{
+	struct emac_instance *dev = param;
+
+	return !(dev->rx_desc[dev->rx_slot].ctrl & MAL_RX_CTRL_EMPTY);
+}
+
+/* NAPI poll context */
+static int emac_peek_rx_sg(void *param)
+{
+	struct emac_instance *dev = param;
+
+	int slot = dev->rx_slot;
+	while (1) {
+		u16 ctrl = dev->rx_desc[slot].ctrl;
+		if (ctrl & MAL_RX_CTRL_EMPTY)
+			return 0;
+		else if (ctrl & MAL_RX_CTRL_LAST)
+			return 1;
+
+		slot = (slot + 1) % NUM_RX_BUFF;
+
+		/* I'm just being paranoid here :) */
+		if (unlikely(slot == dev->rx_slot))
+			return 0;
+	}
+}
+
+/* Hard IRQ */
+static void emac_rxde(void *param)
+{
+	struct emac_instance *dev = param;
+
+	++dev->estats.rx_stopped;
+	emac_rx_disable_async(dev);
+}
+
+/* Hard IRQ */
+static irqreturn_t emac_irq(int irq, void *dev_instance)
+{
+	struct emac_instance *dev = dev_instance;
+	struct emac_regs __iomem *p = dev->emacp;
+	struct emac_error_stats *st = &dev->estats;
+	u32 isr;
+
+	spin_lock(&dev->lock);
+
+	isr = in_be32(&p->isr);
+	out_be32(&p->isr, isr);
+
+	DBG(dev, "isr = %08x" NL, isr);
+
+	if (isr & EMAC4_ISR_TXPE)
+		++st->tx_parity;
+	if (isr & EMAC4_ISR_RXPE)
+		++st->rx_parity;
+	if (isr & EMAC4_ISR_TXUE)
+		++st->tx_underrun;
+	if (isr & EMAC4_ISR_RXOE)
+		++st->rx_fifo_overrun;
+	if (isr & EMAC_ISR_OVR)
+		++st->rx_overrun;
+	if (isr & EMAC_ISR_BP)
+		++st->rx_bad_packet;
+	if (isr & EMAC_ISR_RP)
+		++st->rx_runt_packet;
+	if (isr & EMAC_ISR_SE)
+		++st->rx_short_event;
+	if (isr & EMAC_ISR_ALE)
+		++st->rx_alignment_error;
+	if (isr & EMAC_ISR_BFCS)
+		++st->rx_bad_fcs;
+	if (isr & EMAC_ISR_PTLE)
+		++st->rx_packet_too_long;
+	if (isr & EMAC_ISR_ORE)
+		++st->rx_out_of_range;
+	if (isr & EMAC_ISR_IRE)
+		++st->rx_in_range;
+	if (isr & EMAC_ISR_SQE)
+		++st->tx_sqe;
+	if (isr & EMAC_ISR_TE)
+		++st->tx_errors;
+
+	spin_unlock(&dev->lock);
+
+	return IRQ_HANDLED;
+}
+
+static struct net_device_stats *emac_stats(struct net_device *ndev)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+	struct emac_stats *st = &dev->stats;
+	struct emac_error_stats *est = &dev->estats;
+	struct net_device_stats *nst = &dev->nstats;
+	unsigned long flags;
+
+	DBG2(dev, "stats" NL);
+
+	/* Compute "legacy" statistics */
+	spin_lock_irqsave(&dev->lock, flags);
+	nst->rx_packets = (unsigned long)st->rx_packets;
+	nst->rx_bytes = (unsigned long)st->rx_bytes;
+	nst->tx_packets = (unsigned long)st->tx_packets;
+	nst->tx_bytes = (unsigned long)st->tx_bytes;
+	nst->rx_dropped = (unsigned long)(est->rx_dropped_oom +
+					  est->rx_dropped_error +
+					  est->rx_dropped_resize +
+					  est->rx_dropped_mtu);
+	nst->tx_dropped = (unsigned long)est->tx_dropped;
+
+	nst->rx_errors = (unsigned long)est->rx_bd_errors;
+	nst->rx_fifo_errors = (unsigned long)(est->rx_bd_overrun +
+					      est->rx_fifo_overrun +
+					      est->rx_overrun);
+	nst->rx_frame_errors = (unsigned long)(est->rx_bd_alignment_error +
+					       est->rx_alignment_error);
+	nst->rx_crc_errors = (unsigned long)(est->rx_bd_bad_fcs +
+					     est->rx_bad_fcs);
+	nst->rx_length_errors = (unsigned long)(est->rx_bd_runt_packet +
+						est->rx_bd_short_event +
+						est->rx_bd_packet_too_long +
+						est->rx_bd_out_of_range +
+						est->rx_bd_in_range +
+						est->rx_runt_packet +
+						est->rx_short_event +
+						est->rx_packet_too_long +
+						est->rx_out_of_range +
+						est->rx_in_range);
+
+	nst->tx_errors = (unsigned long)(est->tx_bd_errors + est->tx_errors);
+	nst->tx_fifo_errors = (unsigned long)(est->tx_bd_underrun +
+					      est->tx_underrun);
+	nst->tx_carrier_errors = (unsigned long)est->tx_bd_carrier_loss;
+	nst->collisions = (unsigned long)(est->tx_bd_excessive_deferral +
+					  est->tx_bd_excessive_collisions +
+					  est->tx_bd_late_collision +
+					  est->tx_bd_multple_collisions);
+	spin_unlock_irqrestore(&dev->lock, flags);
+	return nst;
+}
+
+static struct mal_commac_ops emac_commac_ops = {
+	.poll_tx = &emac_poll_tx,
+	.poll_rx = &emac_poll_rx,
+	.peek_rx = &emac_peek_rx,
+	.rxde = &emac_rxde,
+};
+
+static struct mal_commac_ops emac_commac_sg_ops = {
+	.poll_tx = &emac_poll_tx,
+	.poll_rx = &emac_poll_rx,
+	.peek_rx = &emac_peek_rx_sg,
+	.rxde = &emac_rxde,
+};
+
+/* Ethtool support */
+static int emac_ethtool_get_settings(struct net_device *ndev,
+				     struct ethtool_cmd *cmd)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+
+	cmd->supported = dev->phy.features;
+	cmd->port = PORT_MII;
+	cmd->phy_address = dev->phy.address;
+	cmd->transceiver =
+	    dev->phy.address >= 0 ? XCVR_EXTERNAL : XCVR_INTERNAL;
+
+	mutex_lock(&dev->link_lock);
+	cmd->advertising = dev->phy.advertising;
+	cmd->autoneg = dev->phy.autoneg;
+	cmd->speed = dev->phy.speed;
+	cmd->duplex = dev->phy.duplex;
+	mutex_unlock(&dev->link_lock);
+
+	return 0;
+}
+
+static int emac_ethtool_set_settings(struct net_device *ndev,
+				     struct ethtool_cmd *cmd)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+	u32 f = dev->phy.features;
+
+	DBG(dev, "set_settings(%d, %d, %d, 0x%08x)" NL,
+	    cmd->autoneg, cmd->speed, cmd->duplex, cmd->advertising);
+
+	/* Basic sanity checks */
+	if (dev->phy.address < 0)
+		return -EOPNOTSUPP;
+	if (cmd->autoneg != AUTONEG_ENABLE && cmd->autoneg != AUTONEG_DISABLE)
+		return -EINVAL;
+	if (cmd->autoneg == AUTONEG_ENABLE && cmd->advertising == 0)
+		return -EINVAL;
+	if (cmd->duplex != DUPLEX_HALF && cmd->duplex != DUPLEX_FULL)
+		return -EINVAL;
+
+	if (cmd->autoneg == AUTONEG_DISABLE) {
+		switch (cmd->speed) {
+		case SPEED_10:
+			if (cmd->duplex == DUPLEX_HALF &&
+			    !(f & SUPPORTED_10baseT_Half))
+				return -EINVAL;
+			if (cmd->duplex == DUPLEX_FULL &&
+			    !(f & SUPPORTED_10baseT_Full))
+				return -EINVAL;
+			break;
+		case SPEED_100:
+			if (cmd->duplex == DUPLEX_HALF &&
+			    !(f & SUPPORTED_100baseT_Half))
+				return -EINVAL;
+			if (cmd->duplex == DUPLEX_FULL &&
+			    !(f & SUPPORTED_100baseT_Full))
+				return -EINVAL;
+			break;
+		case SPEED_1000:
+			if (cmd->duplex == DUPLEX_HALF &&
+			    !(f & SUPPORTED_1000baseT_Half))
+				return -EINVAL;
+			if (cmd->duplex == DUPLEX_FULL &&
+			    !(f & SUPPORTED_1000baseT_Full))
+				return -EINVAL;
+			break;
+		default:
+			return -EINVAL;
+		}
+
+		mutex_lock(&dev->link_lock);
+		dev->phy.def->ops->setup_forced(&dev->phy, cmd->speed,
+						cmd->duplex);
+		mutex_unlock(&dev->link_lock);
+
+	} else {
+		if (!(f & SUPPORTED_Autoneg))
+			return -EINVAL;
+
+		mutex_lock(&dev->link_lock);
+		dev->phy.def->ops->setup_aneg(&dev->phy,
+					      (cmd->advertising & f) |
+					      (dev->phy.advertising &
+					       (ADVERTISED_Pause |
+						ADVERTISED_Asym_Pause)));
+		mutex_unlock(&dev->link_lock);
+	}
+	emac_force_link_update(dev);
+
+	return 0;
+}
+
+static void emac_ethtool_get_ringparam(struct net_device *ndev,
+				       struct ethtool_ringparam *rp)
+{
+	rp->rx_max_pending = rp->rx_pending = NUM_RX_BUFF;
+	rp->tx_max_pending = rp->tx_pending = NUM_TX_BUFF;
+}
+
+static void emac_ethtool_get_pauseparam(struct net_device *ndev,
+					struct ethtool_pauseparam *pp)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+
+	mutex_lock(&dev->link_lock);
+	if ((dev->phy.features & SUPPORTED_Autoneg) &&
+	    (dev->phy.advertising & (ADVERTISED_Pause | ADVERTISED_Asym_Pause)))
+		pp->autoneg = 1;
+
+	if (dev->phy.duplex == DUPLEX_FULL) {
+		if (dev->phy.pause)
+			pp->rx_pause = pp->tx_pause = 1;
+		else if (dev->phy.asym_pause)
+			pp->tx_pause = 1;
+	}
+	mutex_unlock(&dev->link_lock);
+}
+
+static int emac_get_regs_len(struct emac_instance *dev)
+{
+		return sizeof(struct emac_ethtool_regs_subhdr) +
+			sizeof(struct emac_regs);
+}
+
+static int emac_ethtool_get_regs_len(struct net_device *ndev)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+	int size;
+
+	size = sizeof(struct emac_ethtool_regs_hdr) +
+		emac_get_regs_len(dev) + mal_get_regs_len(dev->mal);
+	if (emac_has_feature(dev, EMAC_FTR_HAS_ZMII))
+		size += zmii_get_regs_len(dev->zmii_dev);
+	if (emac_has_feature(dev, EMAC_FTR_HAS_RGMII))
+		size += rgmii_get_regs_len(dev->rgmii_dev);
+	if (emac_has_feature(dev, EMAC_FTR_HAS_TAH))
+		size += tah_get_regs_len(dev->tah_dev);
+
+	return size;
+}
+
+static void *emac_dump_regs(struct emac_instance *dev, void *buf)
+{
+	struct emac_ethtool_regs_subhdr *hdr = buf;
+
+	hdr->index = dev->cell_index;
+	if (emac_has_feature(dev, EMAC_FTR_EMAC4SYNC)) {
+		hdr->version = EMAC4SYNC_ETHTOOL_REGS_VER;
+	} else if (emac_has_feature(dev, EMAC_FTR_EMAC4)) {
+		hdr->version = EMAC4_ETHTOOL_REGS_VER;
+	} else {
+		hdr->version = EMAC_ETHTOOL_REGS_VER;
+	}
+	memcpy_fromio(hdr + 1, dev->emacp, sizeof(struct emac_regs));
+	return (void *)(hdr + 1) + sizeof(struct emac_regs);
+}
+
+static void emac_ethtool_get_regs(struct net_device *ndev,
+				  struct ethtool_regs *regs, void *buf)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+	struct emac_ethtool_regs_hdr *hdr = buf;
+
+	hdr->components = 0;
+	buf = hdr + 1;
+
+	buf = mal_dump_regs(dev->mal, buf);
+	buf = emac_dump_regs(dev, buf);
+	if (emac_has_feature(dev, EMAC_FTR_HAS_ZMII)) {
+		hdr->components |= EMAC_ETHTOOL_REGS_ZMII;
+		buf = zmii_dump_regs(dev->zmii_dev, buf);
+	}
+	if (emac_has_feature(dev, EMAC_FTR_HAS_RGMII)) {
+		hdr->components |= EMAC_ETHTOOL_REGS_RGMII;
+		buf = rgmii_dump_regs(dev->rgmii_dev, buf);
+	}
+	if (emac_has_feature(dev, EMAC_FTR_HAS_TAH)) {
+		hdr->components |= EMAC_ETHTOOL_REGS_TAH;
+		buf = tah_dump_regs(dev->tah_dev, buf);
+	}
+}
+
+static int emac_ethtool_nway_reset(struct net_device *ndev)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+	int res = 0;
+
+	DBG(dev, "nway_reset" NL);
+
+	if (dev->phy.address < 0)
+		return -EOPNOTSUPP;
+
+	mutex_lock(&dev->link_lock);
+	if (!dev->phy.autoneg) {
+		res = -EINVAL;
+		goto out;
+	}
+
+	dev->phy.def->ops->setup_aneg(&dev->phy, dev->phy.advertising);
+ out:
+	mutex_unlock(&dev->link_lock);
+	emac_force_link_update(dev);
+	return res;
+}
+
+static int emac_ethtool_get_sset_count(struct net_device *ndev, int stringset)
+{
+	if (stringset == ETH_SS_STATS)
+		return EMAC_ETHTOOL_STATS_COUNT;
+	else
+		return -EINVAL;
+}
+
+static void emac_ethtool_get_strings(struct net_device *ndev, u32 stringset,
+				     u8 * buf)
+{
+	if (stringset == ETH_SS_STATS)
+		memcpy(buf, &emac_stats_keys, sizeof(emac_stats_keys));
+}
+
+static void emac_ethtool_get_ethtool_stats(struct net_device *ndev,
+					   struct ethtool_stats *estats,
+					   u64 * tmp_stats)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+
+	memcpy(tmp_stats, &dev->stats, sizeof(dev->stats));
+	tmp_stats += sizeof(dev->stats) / sizeof(u64);
+	memcpy(tmp_stats, &dev->estats, sizeof(dev->estats));
+}
+
+static void emac_ethtool_get_drvinfo(struct net_device *ndev,
+				     struct ethtool_drvinfo *info)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+
+	strlcpy(info->driver, "ibm_emac", sizeof(info->driver));
+	strlcpy(info->version, DRV_VERSION, sizeof(info->version));
+	snprintf(info->bus_info, sizeof(info->bus_info), "PPC 4xx EMAC-%d %s",
+		 dev->cell_index, dev->ofdev->dev.of_node->full_name);
+	info->regdump_len = emac_ethtool_get_regs_len(ndev);
+}
+
+static const struct ethtool_ops emac_ethtool_ops = {
+	.get_settings = emac_ethtool_get_settings,
+	.set_settings = emac_ethtool_set_settings,
+	.get_drvinfo = emac_ethtool_get_drvinfo,
+
+	.get_regs_len = emac_ethtool_get_regs_len,
+	.get_regs = emac_ethtool_get_regs,
+
+	.nway_reset = emac_ethtool_nway_reset,
+
+	.get_ringparam = emac_ethtool_get_ringparam,
+	.get_pauseparam = emac_ethtool_get_pauseparam,
+
+	.get_strings = emac_ethtool_get_strings,
+	.get_sset_count = emac_ethtool_get_sset_count,
+	.get_ethtool_stats = emac_ethtool_get_ethtool_stats,
+
+	.get_link = ethtool_op_get_link,
+};
+
+static int emac_ioctl(struct net_device *ndev, struct ifreq *rq, int cmd)
+{
+	struct emac_instance *dev = netdev_priv(ndev);
+	struct mii_ioctl_data *data = if_mii(rq);
+
+	DBG(dev, "ioctl %08x" NL, cmd);
+
+	if (dev->phy.address < 0)
+		return -EOPNOTSUPP;
+
+	switch (cmd) {
+	case SIOCGMIIPHY:
+		data->phy_id = dev->phy.address;
+		/* Fall through */
+	case SIOCGMIIREG:
+		data->val_out = emac_mdio_read(ndev, dev->phy.address,
+					       data->reg_num);
+		return 0;
+
+	case SIOCSMIIREG:
+		emac_mdio_write(ndev, dev->phy.address, data->reg_num,
+				data->val_in);
+		return 0;
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
+struct emac_depentry {
+	u32			phandle;
+	struct device_node	*node;
+	struct platform_device	*ofdev;
+	void			*drvdata;
+};
+
+#define	EMAC_DEP_MAL_IDX	0
+#define	EMAC_DEP_ZMII_IDX	1
+#define	EMAC_DEP_RGMII_IDX	2
+#define	EMAC_DEP_TAH_IDX	3
+#define	EMAC_DEP_MDIO_IDX	4
+#define	EMAC_DEP_PREV_IDX	5
+#define	EMAC_DEP_COUNT		6
+
+static int emac_check_deps(struct emac_instance *dev,
+			   struct emac_depentry *deps)
+{
+	int i, there = 0;
+	struct device_node *np;
+
+	for (i = 0; i < EMAC_DEP_COUNT; i++) {
+		/* no dependency on that item, allright */
+		if (deps[i].phandle == 0) {
+			there++;
+			continue;
+		}
+		/* special case for blist as the dependency might go away */
+		if (i == EMAC_DEP_PREV_IDX) {
+			np = *(dev->blist - 1);
+			if (np == NULL) {
+				deps[i].phandle = 0;
+				there++;
+				continue;
+			}
+			if (deps[i].node == NULL)
+				deps[i].node = of_node_get(np);
+		}
+		if (deps[i].node == NULL)
+			deps[i].node = of_find_node_by_phandle(deps[i].phandle);
+		if (deps[i].node == NULL)
+			continue;
+		if (deps[i].ofdev == NULL)
+			deps[i].ofdev = of_find_device_by_node(deps[i].node);
+		if (deps[i].ofdev == NULL)
+			continue;
+		if (deps[i].drvdata == NULL)
+			deps[i].drvdata = platform_get_drvdata(deps[i].ofdev);
+		if (deps[i].drvdata != NULL)
+			there++;
+	}
+	return there == EMAC_DEP_COUNT;
+}
+
+static void emac_put_deps(struct emac_instance *dev)
+{
+	of_dev_put(dev->mal_dev);
+	of_dev_put(dev->zmii_dev);
+	of_dev_put(dev->rgmii_dev);
+	of_dev_put(dev->mdio_dev);
+	of_dev_put(dev->tah_dev);
+}
+
+static int emac_of_bus_notify(struct notifier_block *nb, unsigned long action,
+			      void *data)
+{
+	/* We are only intereted in device addition */
+	if (action == BUS_NOTIFY_BOUND_DRIVER)
+		wake_up_all(&emac_probe_wait);
+	return 0;
+}
+
+static struct notifier_block emac_of_bus_notifier = {
+	.notifier_call = emac_of_bus_notify
+};
+
+static int emac_wait_deps(struct emac_instance *dev)
+{
+	struct emac_depentry deps[EMAC_DEP_COUNT];
+	int i, err;
+
+	memset(&deps, 0, sizeof(deps));
+
+	deps[EMAC_DEP_MAL_IDX].phandle = dev->mal_ph;
+	deps[EMAC_DEP_ZMII_IDX].phandle = dev->zmii_ph;
+	deps[EMAC_DEP_RGMII_IDX].phandle = dev->rgmii_ph;
+	if (dev->tah_ph)
+		deps[EMAC_DEP_TAH_IDX].phandle = dev->tah_ph;
+	if (dev->mdio_ph)
+		deps[EMAC_DEP_MDIO_IDX].phandle = dev->mdio_ph;
+	if (dev->blist && dev->blist > emac_boot_list)
+		deps[EMAC_DEP_PREV_IDX].phandle = 0xffffffffu;
+	bus_register_notifier(&platform_bus_type, &emac_of_bus_notifier);
+	wait_event_timeout(emac_probe_wait,
+			   emac_check_deps(dev, deps),
+			   EMAC_PROBE_DEP_TIMEOUT);
+	bus_unregister_notifier(&platform_bus_type, &emac_of_bus_notifier);
+	err = emac_check_deps(dev, deps) ? 0 : -ENODEV;
+	for (i = 0; i < EMAC_DEP_COUNT; i++) {
+		of_node_put(deps[i].node);
+		if (err)
+			of_dev_put(deps[i].ofdev);
+	}
+	if (err == 0) {
+		dev->mal_dev = deps[EMAC_DEP_MAL_IDX].ofdev;
+		dev->zmii_dev = deps[EMAC_DEP_ZMII_IDX].ofdev;
+		dev->rgmii_dev = deps[EMAC_DEP_RGMII_IDX].ofdev;
+		dev->tah_dev = deps[EMAC_DEP_TAH_IDX].ofdev;
+		dev->mdio_dev = deps[EMAC_DEP_MDIO_IDX].ofdev;
+	}
+	of_dev_put(deps[EMAC_DEP_PREV_IDX].ofdev);
+	return err;
+}
+
+static int emac_read_uint_prop(struct device_node *np, const char *name,
+			       u32 *val, int fatal)
+{
+	int len;
+	const u32 *prop = of_get_property(np, name, &len);
+	if (prop == NULL || len < sizeof(u32)) {
+		if (fatal)
+			printk(KERN_ERR "%s: missing %s property\n",
+			       np->full_name, name);
+		return -ENODEV;
+	}
+	*val = *prop;
+	return 0;
+}
+
+static int emac_init_phy(struct emac_instance *dev)
+{
+	struct device_node *np = dev->ofdev->dev.of_node;
+	struct net_device *ndev = dev->ndev;
+	u32 phy_map, adv;
+	int i;
+
+	dev->phy.dev = ndev;
+	dev->phy.mode = dev->phy_mode;
+
+	/* PHY-less configuration.
+	 * XXX I probably should move these settings to the dev tree
+	 */
+	if (dev->phy_address == 0xffffffff && dev->phy_map == 0xffffffff) {
+		emac_reset(dev);
+
+		/* PHY-less configuration.
+		 * XXX I probably should move these settings to the dev tree
+		 */
+		dev->phy.address = -1;
+		dev->phy.features = SUPPORTED_MII;
+		if (emac_phy_supports_gige(dev->phy_mode))
+			dev->phy.features |= SUPPORTED_1000baseT_Full;
+		else
+			dev->phy.features |= SUPPORTED_100baseT_Full;
+		dev->phy.pause = 1;
+
+		return 0;
+	}
+
+	mutex_lock(&emac_phy_map_lock);
+	phy_map = dev->phy_map | busy_phy_map;
+
+	DBG(dev, "PHY maps %08x %08x" NL, dev->phy_map, busy_phy_map);
+
+	dev->phy.mdio_read = emac_mdio_read;
+	dev->phy.mdio_write = emac_mdio_write;
+
+	/* Enable internal clock source */
+#ifdef CONFIG_PPC_DCR_NATIVE
+	if (emac_has_feature(dev, EMAC_FTR_440GX_PHY_CLK_FIX))
+		dcri_clrset(SDR0, SDR0_MFR, 0, SDR0_MFR_ECS);
+#endif
+	/* PHY clock workaround */
+	emac_rx_clk_tx(dev);
+
+	/* Enable internal clock source on 440GX*/
+#ifdef CONFIG_PPC_DCR_NATIVE
+	if (emac_has_feature(dev, EMAC_FTR_440GX_PHY_CLK_FIX))
+		dcri_clrset(SDR0, SDR0_MFR, 0, SDR0_MFR_ECS);
+#endif
+	/* Configure EMAC with defaults so we can at least use MDIO
+	 * This is needed mostly for 440GX
+	 */
+	if (emac_phy_gpcs(dev->phy.mode)) {
+		/* XXX
+		 * Make GPCS PHY address equal to EMAC index.
+		 * We probably should take into account busy_phy_map
+		 * and/or phy_map here.
+		 *
+		 * Note that the busy_phy_map is currently global
+		 * while it should probably be per-ASIC...
+		 */
+		dev->phy.gpcs_address = dev->gpcs_address;
+		if (dev->phy.gpcs_address == 0xffffffff)
+			dev->phy.address = dev->cell_index;
+	}
+
+	emac_configure(dev);
+
+	if (dev->phy_address != 0xffffffff)
+		phy_map = ~(1 << dev->phy_address);
+
+	for (i = 0; i < 0x20; phy_map >>= 1, ++i)
+		if (!(phy_map & 1)) {
+			int r;
+			busy_phy_map |= 1 << i;
+
+			/* Quick check if there is a PHY at the address */
+			r = emac_mdio_read(dev->ndev, i, MII_BMCR);
+			if (r == 0xffff || r < 0)
+				continue;
+			if (!emac_mii_phy_probe(&dev->phy, i))
+				break;
+		}
+
+	/* Enable external clock source */
+#ifdef CONFIG_PPC_DCR_NATIVE
+	if (emac_has_feature(dev, EMAC_FTR_440GX_PHY_CLK_FIX))
+		dcri_clrset(SDR0, SDR0_MFR, SDR0_MFR_ECS, 0);
+#endif
+	mutex_unlock(&emac_phy_map_lock);
+	if (i == 0x20) {
+		printk(KERN_WARNING "%s: can't find PHY!\n", np->full_name);
+		return -ENXIO;
+	}
+
+	/* Init PHY */
+	if (dev->phy.def->ops->init)
+		dev->phy.def->ops->init(&dev->phy);
+
+	/* Disable any PHY features not supported by the platform */
+	dev->phy.def->features &= ~dev->phy_feat_exc;
+	dev->phy.features &= ~dev->phy_feat_exc;
+
+	/* Setup initial link parameters */
+	if (dev->phy.features & SUPPORTED_Autoneg) {
+		adv = dev->phy.features;
+		if (!emac_has_feature(dev, EMAC_FTR_NO_FLOW_CONTROL_40x))
+			adv |= ADVERTISED_Pause | ADVERTISED_Asym_Pause;
+		/* Restart autonegotiation */
+		dev->phy.def->ops->setup_aneg(&dev->phy, adv);
+	} else {
+		u32 f = dev->phy.def->features;
+		int speed = SPEED_10, fd = DUPLEX_HALF;
+
+		/* Select highest supported speed/duplex */
+		if (f & SUPPORTED_1000baseT_Full) {
+			speed = SPEED_1000;
+			fd = DUPLEX_FULL;
+		} else if (f & SUPPORTED_1000baseT_Half)
+			speed = SPEED_1000;
+		else if (f & SUPPORTED_100baseT_Full) {
+			speed = SPEED_100;
+			fd = DUPLEX_FULL;
+		} else if (f & SUPPORTED_100baseT_Half)
+			speed = SPEED_100;
+		else if (f & SUPPORTED_10baseT_Full)
+			fd = DUPLEX_FULL;
+
+		/* Force link parameters */
+		dev->phy.def->ops->setup_forced(&dev->phy, speed, fd);
+	}
+	return 0;
+}
+
+static int emac_init_config(struct emac_instance *dev)
+{
+	struct device_node *np = dev->ofdev->dev.of_node;
+	const void *p;
+
+	/* Read config from device-tree */
+	if (emac_read_uint_prop(np, "mal-device", &dev->mal_ph, 1))
+		return -ENXIO;
+	if (emac_read_uint_prop(np, "mal-tx-channel", &dev->mal_tx_chan, 1))
+		return -ENXIO;
+	if (emac_read_uint_prop(np, "mal-rx-channel", &dev->mal_rx_chan, 1))
+		return -ENXIO;
+	if (emac_read_uint_prop(np, "cell-index", &dev->cell_index, 1))
+		return -ENXIO;
+	if (emac_read_uint_prop(np, "max-frame-size", &dev->max_mtu, 0))
+		dev->max_mtu = 1500;
+	if (emac_read_uint_prop(np, "rx-fifo-size", &dev->rx_fifo_size, 0))
+		dev->rx_fifo_size = 2048;
+	if (emac_read_uint_prop(np, "tx-fifo-size", &dev->tx_fifo_size, 0))
+		dev->tx_fifo_size = 2048;
+	if (emac_read_uint_prop(np, "rx-fifo-size-gige", &dev->rx_fifo_size_gige, 0))
+		dev->rx_fifo_size_gige = dev->rx_fifo_size;
+	if (emac_read_uint_prop(np, "tx-fifo-size-gige", &dev->tx_fifo_size_gige, 0))
+		dev->tx_fifo_size_gige = dev->tx_fifo_size;
+	if (emac_read_uint_prop(np, "phy-address", &dev->phy_address, 0))
+		dev->phy_address = 0xffffffff;
+	if (emac_read_uint_prop(np, "phy-map", &dev->phy_map, 0))
+		dev->phy_map = 0xffffffff;
+	if (emac_read_uint_prop(np, "gpcs-address", &dev->gpcs_address, 0))
+		dev->gpcs_address = 0xffffffff;
+	if (emac_read_uint_prop(np->parent, "clock-frequency", &dev->opb_bus_freq, 1))
+		return -ENXIO;
+	if (emac_read_uint_prop(np, "tah-device", &dev->tah_ph, 0))
+		dev->tah_ph = 0;
+	if (emac_read_uint_prop(np, "tah-channel", &dev->tah_port, 0))
+		dev->tah_port = 0;
+	if (emac_read_uint_prop(np, "mdio-device", &dev->mdio_ph, 0))
+		dev->mdio_ph = 0;
+	if (emac_read_uint_prop(np, "zmii-device", &dev->zmii_ph, 0))
+		dev->zmii_ph = 0;
+	if (emac_read_uint_prop(np, "zmii-channel", &dev->zmii_port, 0))
+		dev->zmii_port = 0xffffffff;
+	if (emac_read_uint_prop(np, "rgmii-device", &dev->rgmii_ph, 0))
+		dev->rgmii_ph = 0;
+	if (emac_read_uint_prop(np, "rgmii-channel", &dev->rgmii_port, 0))
+		dev->rgmii_port = 0xffffffff;
+	if (emac_read_uint_prop(np, "fifo-entry-size", &dev->fifo_entry_size, 0))
+		dev->fifo_entry_size = 16;
+	if (emac_read_uint_prop(np, "mal-burst-size", &dev->mal_burst_size, 0))
+		dev->mal_burst_size = 256;
+
+	/* PHY mode needs some decoding */
+	dev->phy_mode = of_get_phy_mode(np);
+	if (dev->phy_mode < 0)
+		dev->phy_mode = PHY_MODE_NA;
+
+	/* Check EMAC version */
+	if (of_device_is_compatible(np, "ibm,emac4sync")) {
+		dev->features |= (EMAC_FTR_EMAC4 | EMAC_FTR_EMAC4SYNC);
+		if (of_device_is_compatible(np, "ibm,emac-460ex") ||
+		    of_device_is_compatible(np, "ibm,emac-460gt"))
+			dev->features |= EMAC_FTR_460EX_PHY_CLK_FIX;
+		if (of_device_is_compatible(np, "ibm,emac-405ex") ||
+		    of_device_is_compatible(np, "ibm,emac-405exr"))
+			dev->features |= EMAC_FTR_440EP_PHY_CLK_FIX;
+		if (of_device_is_compatible(np, "ibm,emac-apm821xx")) {
+			dev->features |= (EMAC_APM821XX_REQ_JUMBO_FRAME_SIZE |
+					  EMAC_FTR_APM821XX_NO_HALF_DUPLEX |
+					  EMAC_FTR_460EX_PHY_CLK_FIX);
+		}
+	} else if (of_device_is_compatible(np, "ibm,emac4")) {
+		dev->features |= EMAC_FTR_EMAC4;
+		if (of_device_is_compatible(np, "ibm,emac-440gx"))
+			dev->features |= EMAC_FTR_440GX_PHY_CLK_FIX;
+	} else {
+		if (of_device_is_compatible(np, "ibm,emac-440ep") ||
+		    of_device_is_compatible(np, "ibm,emac-440gr"))
+			dev->features |= EMAC_FTR_440EP_PHY_CLK_FIX;
+		if (of_device_is_compatible(np, "ibm,emac-405ez")) {
+#ifdef CONFIG_IBM_EMAC_NO_FLOW_CTRL
+			dev->features |= EMAC_FTR_NO_FLOW_CONTROL_40x;
+#else
+			printk(KERN_ERR "%s: Flow control not disabled!\n",
+					np->full_name);
+			return -ENXIO;
+#endif
+		}
+
+	}
+
+	/* Fixup some feature bits based on the device tree */
+	if (of_get_property(np, "has-inverted-stacr-oc", NULL))
+		dev->features |= EMAC_FTR_STACR_OC_INVERT;
+	if (of_get_property(np, "has-new-stacr-staopc", NULL))
+		dev->features |= EMAC_FTR_HAS_NEW_STACR;
+
+	/* CAB lacks the appropriate properties */
+	if (of_device_is_compatible(np, "ibm,emac-axon"))
+		dev->features |= EMAC_FTR_HAS_NEW_STACR |
+			EMAC_FTR_STACR_OC_INVERT;
+
+	/* Enable TAH/ZMII/RGMII features as found */
+	if (dev->tah_ph != 0) {
+#ifdef CONFIG_IBM_EMAC_TAH
+		dev->features |= EMAC_FTR_HAS_TAH;
+#else
+		printk(KERN_ERR "%s: TAH support not enabled !\n",
+		       np->full_name);
+		return -ENXIO;
+#endif
+	}
+
+	if (dev->zmii_ph != 0) {
+#ifdef CONFIG_IBM_EMAC_ZMII
+		dev->features |= EMAC_FTR_HAS_ZMII;
+#else
+		printk(KERN_ERR "%s: ZMII support not enabled !\n",
+		       np->full_name);
+		return -ENXIO;
+#endif
+	}
+
+	if (dev->rgmii_ph != 0) {
+#ifdef CONFIG_IBM_EMAC_RGMII
+		dev->features |= EMAC_FTR_HAS_RGMII;
+#else
+		printk(KERN_ERR "%s: RGMII support not enabled !\n",
+		       np->full_name);
+		return -ENXIO;
+#endif
+	}
+
+	/* Read MAC-address */
+	p = of_get_property(np, "local-mac-address", NULL);
+	if (p == NULL) {
+		printk(KERN_ERR "%s: Can't find local-mac-address property\n",
+		       np->full_name);
+		return -ENXIO;
+	}
+	memcpy(dev->ndev->dev_addr, p, ETH_ALEN);
+
+	/* IAHT and GAHT filter parameterization */
+	if (emac_has_feature(dev, EMAC_FTR_EMAC4SYNC)) {
+		dev->xaht_slots_shift = EMAC4SYNC_XAHT_SLOTS_SHIFT;
+		dev->xaht_width_shift = EMAC4SYNC_XAHT_WIDTH_SHIFT;
+	} else {
+		dev->xaht_slots_shift = EMAC4_XAHT_SLOTS_SHIFT;
+		dev->xaht_width_shift = EMAC4_XAHT_WIDTH_SHIFT;
+	}
+
+	DBG(dev, "features     : 0x%08x / 0x%08x\n", dev->features, EMAC_FTRS_POSSIBLE);
+	DBG(dev, "tx_fifo_size : %d (%d gige)\n", dev->tx_fifo_size, dev->tx_fifo_size_gige);
+	DBG(dev, "rx_fifo_size : %d (%d gige)\n", dev->rx_fifo_size, dev->rx_fifo_size_gige);
+	DBG(dev, "max_mtu      : %d\n", dev->max_mtu);
+	DBG(dev, "OPB freq     : %d\n", dev->opb_bus_freq);
+
+	return 0;
+}
+
+static const struct net_device_ops emac_netdev_ops = {
+	.ndo_open		= emac_open,
+	.ndo_stop		= emac_close,
+	.ndo_get_stats		= emac_stats,
+	.ndo_set_rx_mode	= emac_set_multicast_list,
+	.ndo_do_ioctl		= emac_ioctl,
+	.ndo_tx_timeout		= emac_tx_timeout,
+	.ndo_validate_addr	= eth_validate_addr,
+	.ndo_set_mac_address	= eth_mac_addr,
+	.ndo_start_xmit		= emac_start_xmit,
+	.ndo_change_mtu		= eth_change_mtu,
+};
+
+static const struct net_device_ops emac_gige_netdev_ops = {
+	.ndo_open		= emac_open,
+	.ndo_stop		= emac_close,
+	.ndo_get_stats		= emac_stats,
+	.ndo_set_rx_mode	= emac_set_multicast_list,
+	.ndo_do_ioctl		= emac_ioctl,
+	.ndo_tx_timeout		= emac_tx_timeout,
+	.ndo_validate_addr	= eth_validate_addr,
+	.ndo_set_mac_address	= eth_mac_addr,
+	.ndo_start_xmit		= emac_start_xmit_sg,
+	.ndo_change_mtu		= emac_change_mtu,
+};
+
+static int emac_probe(struct platform_device *ofdev)
+{
+	struct net_device *ndev;
+	struct emac_instance *dev;
+	struct device_node *np = ofdev->dev.of_node;
+	struct device_node **blist = NULL;
+	int err, i;
+
+	/* Skip unused/unwired EMACS.  We leave the check for an unused
+	 * property here for now, but new flat device trees should set a
+	 * status property to "disabled" instead.
+	 */
+	if (of_get_property(np, "unused", NULL) || !of_device_is_available(np))
+		return -ENODEV;
+
+	/* Find ourselves in the bootlist if we are there */
+	for (i = 0; i < EMAC_BOOT_LIST_SIZE; i++)
+		if (emac_boot_list[i] == np)
+			blist = &emac_boot_list[i];
+
+	/* Allocate our net_device structure */
+	err = -ENOMEM;
+	ndev = alloc_etherdev(sizeof(struct emac_instance));
+	if (!ndev)
+		goto err_gone;
+
+	dev = netdev_priv(ndev);
+	dev->ndev = ndev;
+	dev->ofdev = ofdev;
+	dev->blist = blist;
+	SET_NETDEV_DEV(ndev, &ofdev->dev);
+
+	/* Initialize some embedded data structures */
+	mutex_init(&dev->mdio_lock);
+	mutex_init(&dev->link_lock);
+	spin_lock_init(&dev->lock);
+	INIT_WORK(&dev->reset_work, emac_reset_work);
+
+	/* Init various config data based on device-tree */
+	err = emac_init_config(dev);
+	if (err != 0)
+		goto err_free;
+
+	/* Get interrupts. EMAC irq is mandatory, WOL irq is optional */
+	dev->emac_irq = irq_of_parse_and_map(np, 0);
+	dev->wol_irq = irq_of_parse_and_map(np, 1);
+	if (dev->emac_irq == NO_IRQ) {
+		printk(KERN_ERR "%s: Can't map main interrupt\n", np->full_name);
+		goto err_free;
+	}
+	ndev->irq = dev->emac_irq;
+
+	/* Map EMAC regs */
+	if (of_address_to_resource(np, 0, &dev->rsrc_regs)) {
+		printk(KERN_ERR "%s: Can't get registers address\n",
+		       np->full_name);
+		goto err_irq_unmap;
+	}
+	// TODO : request_mem_region
+	dev->emacp = ioremap(dev->rsrc_regs.start,
+			     resource_size(&dev->rsrc_regs));
+	if (dev->emacp == NULL) {
+		printk(KERN_ERR "%s: Can't map device registers!\n",
+		       np->full_name);
+		err = -ENOMEM;
+		goto err_irq_unmap;
+	}
+
+	/* Wait for dependent devices */
+	err = emac_wait_deps(dev);
+	if (err) {
+		printk(KERN_ERR
+		       "%s: Timeout waiting for dependent devices\n",
+		       np->full_name);
+		/*  display more info about what's missing ? */
+		goto err_reg_unmap;
+	}
+	dev->mal = platform_get_drvdata(dev->mal_dev);
+	if (dev->mdio_dev != NULL)
+		dev->mdio_instance = platform_get_drvdata(dev->mdio_dev);
+
+	/* Register with MAL */
+	dev->commac.ops = &emac_commac_ops;
+	dev->commac.dev = dev;
+	dev->commac.tx_chan_mask = MAL_CHAN_MASK(dev->mal_tx_chan);
+	dev->commac.rx_chan_mask = MAL_CHAN_MASK(dev->mal_rx_chan);
+	err = mal_register_commac(dev->mal, &dev->commac);
+	if (err) {
+		printk(KERN_ERR "%s: failed to register with mal %s!\n",
+		       np->full_name, dev->mal_dev->dev.of_node->full_name);
+		goto err_rel_deps;
+	}
+	dev->rx_skb_size = emac_rx_skb_size(ndev->mtu);
+	dev->rx_sync_size = emac_rx_sync_size(ndev->mtu);
+
+	/* Get pointers to BD rings */
+	dev->tx_desc =
+	    dev->mal->bd_virt + mal_tx_bd_offset(dev->mal, dev->mal_tx_chan);
+	dev->rx_desc =
+	    dev->mal->bd_virt + mal_rx_bd_offset(dev->mal, dev->mal_rx_chan);
+
+	DBG(dev, "tx_desc %p" NL, dev->tx_desc);
+	DBG(dev, "rx_desc %p" NL, dev->rx_desc);
+
+	/* Clean rings */
+	memset(dev->tx_desc, 0, NUM_TX_BUFF * sizeof(struct mal_descriptor));
+	memset(dev->rx_desc, 0, NUM_RX_BUFF * sizeof(struct mal_descriptor));
+	memset(dev->tx_skb, 0, NUM_TX_BUFF * sizeof(struct sk_buff *));
+	memset(dev->rx_skb, 0, NUM_RX_BUFF * sizeof(struct sk_buff *));
+
+	/* Attach to ZMII, if needed */
+	if (emac_has_feature(dev, EMAC_FTR_HAS_ZMII) &&
+	    (err = zmii_attach(dev->zmii_dev, dev->zmii_port, &dev->phy_mode)) != 0)
+		goto err_unreg_commac;
+
+	/* Attach to RGMII, if needed */
+	if (emac_has_feature(dev, EMAC_FTR_HAS_RGMII) &&
+	    (err = rgmii_attach(dev->rgmii_dev, dev->rgmii_port, dev->phy_mode)) != 0)
+		goto err_detach_zmii;
+
+	/* Attach to TAH, if needed */
+	if (emac_has_feature(dev, EMAC_FTR_HAS_TAH) &&
+	    (err = tah_attach(dev->tah_dev, dev->tah_port)) != 0)
+		goto err_detach_rgmii;
+
+	/* Set some link defaults before we can find out real parameters */
+	dev->phy.speed = SPEED_100;
+	dev->phy.duplex = DUPLEX_FULL;
+	dev->phy.autoneg = AUTONEG_DISABLE;
+	dev->phy.pause = dev->phy.asym_pause = 0;
+	dev->stop_timeout = STOP_TIMEOUT_100;
+	INIT_DELAYED_WORK(&dev->link_work, emac_link_timer);
+
+	/* Some SoCs like APM821xx does not support Half Duplex mode. */
+	if (emac_has_feature(dev, EMAC_FTR_APM821XX_NO_HALF_DUPLEX)) {
+		dev->phy_feat_exc = (SUPPORTED_1000baseT_Half |
+				     SUPPORTED_100baseT_Half |
+				     SUPPORTED_10baseT_Half);
+	}
+
+	/* Find PHY if any */
+	err = emac_init_phy(dev);
+	if (err != 0)
+		goto err_detach_tah;
+
+	if (dev->tah_dev) {
+		ndev->hw_features = NETIF_F_IP_CSUM | NETIF_F_SG;
+		ndev->features |= ndev->hw_features | NETIF_F_RXCSUM;
+	}
+	ndev->watchdog_timeo = 5 * HZ;
+	if (emac_phy_supports_gige(dev->phy_mode)) {
+		ndev->netdev_ops = &emac_gige_netdev_ops;
+		dev->commac.ops = &emac_commac_sg_ops;
+	} else
+		ndev->netdev_ops = &emac_netdev_ops;
+	ndev->ethtool_ops = &emac_ethtool_ops;
+
+	netif_carrier_off(ndev);
+
+	err = register_netdev(ndev);
+	if (err) {
+		printk(KERN_ERR "%s: failed to register net device (%d)!\n",
+		       np->full_name, err);
+		goto err_detach_tah;
+	}
+
+	/* Set our drvdata last as we don't want them visible until we are
+	 * fully initialized
+	 */
+	wmb();
+	platform_set_drvdata(ofdev, dev);
+
+	/* There's a new kid in town ! Let's tell everybody */
+	wake_up_all(&emac_probe_wait);
+
+
+	printk(KERN_INFO "%s: EMAC-%d %s, MAC %pM\n",
+	       ndev->name, dev->cell_index, np->full_name, ndev->dev_addr);
+
+	if (dev->phy_mode == PHY_MODE_SGMII)
+		printk(KERN_NOTICE "%s: in SGMII mode\n", ndev->name);
+
+	if (dev->phy.address >= 0)
+		printk("%s: found %s PHY (0x%02x)\n", ndev->name,
+		       dev->phy.def->name, dev->phy.address);
+
+	emac_dbg_register(dev);
+
+	/* Life is good */
+	return 0;
+
+	/* I have a bad feeling about this ... */
+
+ err_detach_tah:
+	if (emac_has_feature(dev, EMAC_FTR_HAS_TAH))
+		tah_detach(dev->tah_dev, dev->tah_port);
+ err_detach_rgmii:
+	if (emac_has_feature(dev, EMAC_FTR_HAS_RGMII))
+		rgmii_detach(dev->rgmii_dev, dev->rgmii_port);
+ err_detach_zmii:
+	if (emac_has_feature(dev, EMAC_FTR_HAS_ZMII))
+		zmii_detach(dev->zmii_dev, dev->zmii_port);
+ err_unreg_commac:
+	mal_unregister_commac(dev->mal, &dev->commac);
+ err_rel_deps:
+	emac_put_deps(dev);
+ err_reg_unmap:
+	iounmap(dev->emacp);
+ err_irq_unmap:
+	if (dev->wol_irq != NO_IRQ)
+		irq_dispose_mapping(dev->wol_irq);
+	if (dev->emac_irq != NO_IRQ)
+		irq_dispose_mapping(dev->emac_irq);
+ err_free:
+	free_netdev(ndev);
+ err_gone:
+	/* if we were on the bootlist, remove us as we won't show up and
+	 * wake up all waiters to notify them in case they were waiting
+	 * on us
+	 */
+	if (blist) {
+		*blist = NULL;
+		wake_up_all(&emac_probe_wait);
+	}
+	return err;
+}
+
+static int emac_remove(struct platform_device *ofdev)
+{
+	struct emac_instance *dev = platform_get_drvdata(ofdev);
+
+	DBG(dev, "remove" NL);
+
+	unregister_netdev(dev->ndev);
+
+	cancel_work_sync(&dev->reset_work);
+
+	if (emac_has_feature(dev, EMAC_FTR_HAS_TAH))
+		tah_detach(dev->tah_dev, dev->tah_port);
+	if (emac_has_feature(dev, EMAC_FTR_HAS_RGMII))
+		rgmii_detach(dev->rgmii_dev, dev->rgmii_port);
+	if (emac_has_feature(dev, EMAC_FTR_HAS_ZMII))
+		zmii_detach(dev->zmii_dev, dev->zmii_port);
+
+	busy_phy_map &= ~(1 << dev->phy.address);
+	DBG(dev, "busy_phy_map now %#x" NL, busy_phy_map);
+
+	mal_unregister_commac(dev->mal, &dev->commac);
+	emac_put_deps(dev);
+
+	emac_dbg_unregister(dev);
+	iounmap(dev->emacp);
+
+	if (dev->wol_irq != NO_IRQ)
+		irq_dispose_mapping(dev->wol_irq);
+	if (dev->emac_irq != NO_IRQ)
+		irq_dispose_mapping(dev->emac_irq);
+
+	free_netdev(dev->ndev);
+
+	return 0;
+}
+
+/* XXX Features in here should be replaced by properties... */
+static const struct of_device_id emac_match[] =
+{
+	{
+		.type		= "network",
+		.compatible	= "ibm,emac",
+	},
+	{
+		.type		= "network",
+		.compatible	= "ibm,emac4",
+	},
+	{
+		.type		= "network",
+		.compatible	= "ibm,emac4sync",
+	},
+	{},
+};
+MODULE_DEVICE_TABLE(of, emac_match);
+
+static struct platform_driver emac_driver = {
+	.driver = {
+		.name = "emac",
+		.of_match_table = emac_match,
+	},
+	.probe = emac_probe,
+	.remove = emac_remove,
+};
+
+static void __init emac_make_bootlist(void)
+{
+	struct device_node *np = NULL;
+	int j, max, i = 0, k;
+	int cell_indices[EMAC_BOOT_LIST_SIZE];
+
+	/* Collect EMACs */
+	while((np = of_find_all_nodes(np)) != NULL) {
+		const u32 *idx;
+
+		if (of_match_node(emac_match, np) == NULL)
+			continue;
+		if (of_get_property(np, "unused", NULL))
+			continue;
+		idx = of_get_property(np, "cell-index", NULL);
+		if (idx == NULL)
+			continue;
+		cell_indices[i] = *idx;
+		emac_boot_list[i++] = of_node_get(np);
+		if (i >= EMAC_BOOT_LIST_SIZE) {
+			of_node_put(np);
+			break;
+		}
+	}
+	max = i;
+
+	/* Bubble sort them (doh, what a creative algorithm :-) */
+	for (i = 0; max > 1 && (i < (max - 1)); i++)
+		for (j = i; j < max; j++) {
+			if (cell_indices[i] > cell_indices[j]) {
+				np = emac_boot_list[i];
+				emac_boot_list[i] = emac_boot_list[j];
+				emac_boot_list[j] = np;
+				k = cell_indices[i];
+				cell_indices[i] = cell_indices[j];
+				cell_indices[j] = k;
+			}
+		}
+}
+
+static int __init emac_init(void)
+{
+	int rc;
+
+	printk(KERN_INFO DRV_DESC ", version " DRV_VERSION "\n");
+
+	/* Init debug stuff */
+	emac_init_debug();
+
+	/* Build EMAC boot list */
+	emac_make_bootlist();
+
+	/* Init submodules */
+	rc = mal_init();
+	if (rc)
+		goto err;
+	rc = zmii_init();
+	if (rc)
+		goto err_mal;
+	rc = rgmii_init();
+	if (rc)
+		goto err_zmii;
+	rc = tah_init();
+	if (rc)
+		goto err_rgmii;
+	rc = platform_driver_register(&emac_driver);
+	if (rc)
+		goto err_tah;
+
+	return 0;
+
+ err_tah:
+	tah_exit();
+ err_rgmii:
+	rgmii_exit();
+ err_zmii:
+	zmii_exit();
+ err_mal:
+	mal_exit();
+ err:
+	return rc;
+}
+
+static void __exit emac_exit(void)
+{
+	int i;
+
+	platform_driver_unregister(&emac_driver);
+
+	tah_exit();
+	rgmii_exit();
+	zmii_exit();
+	mal_exit();
+	emac_fini_debug();
+
+	/* Destroy EMAC boot list */
+	for (i = 0; i < EMAC_BOOT_LIST_SIZE; i++)
+		of_node_put(emac_boot_list[i]);
+}
+
+module_init(emac_init);
+module_exit(emac_exit);
diff --git a/drivers/net/ethernet/ibm/emac/core.h b/drivers/net/ethernet/ibm/emac/core.h
new file mode 100644
index 000000000..28df37420
--- /dev/null
+++ b/drivers/net/ethernet/ibm/emac/core.h
@@ -0,0 +1,467 @@
+/*
+ * drivers/net/ethernet/ibm/emac/core.h
+ *
+ * Driver for PowerPC 4xx on-chip ethernet controller.
+ *
+ * Copyright 2007 Benjamin Herrenschmidt, IBM Corp.
+ *                <benh@kernel.crashing.org>
+ *
+ * Based on the arch/ppc version of the driver:
+ *
+ * Copyright (c) 2004, 2005 Zultys Technologies.
+ * Eugene Surovegin <eugene.surovegin@zultys.com> or <ebs@ebshome.net>
+ *
+ * Based on original work by
+ *      Armin Kuster <akuster@mvista.com>
+ * 	Johnnie Peters <jpeters@mvista.com>
+ *      Copyright 2000, 2001 MontaVista Softare Inc.
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ *
+ */
+#ifndef __IBM_NEWEMAC_CORE_H
+#define __IBM_NEWEMAC_CORE_H
+
+#include <linux/module.h>
+#include <linux/list.h>
+#include <linux/kernel.h>
+#include <linux/interrupt.h>
+#include <linux/netdevice.h>
+#include <linux/dma-mapping.h>
+#include <linux/spinlock.h>
+#include <linux/of_platform.h>
+#include <linux/slab.h>
+
+#include <asm/io.h>
+#include <asm/dcr.h>
+
+#include "emac.h"
+#include "phy.h"
+#include "zmii.h"
+#include "rgmii.h"
+#include "mal.h"
+#include "tah.h"
+#include "debug.h"
+
+#define NUM_TX_BUFF			CONFIG_IBM_EMAC_TXB
+#define NUM_RX_BUFF			CONFIG_IBM_EMAC_RXB
+
+/* Simple sanity check */
+#if NUM_TX_BUFF > 256 || NUM_RX_BUFF > 256
+#error Invalid number of buffer descriptors (greater than 256)
+#endif
+
+#define EMAC_MIN_MTU			46
+
+/* Maximum L2 header length (VLAN tagged, no FCS) */
+#define EMAC_MTU_OVERHEAD		(6 * 2 + 2 + 4)
+
+/* RX BD size for the given MTU */
+static inline int emac_rx_size(int mtu)
+{
+	if (mtu > ETH_DATA_LEN)
+		return MAL_MAX_RX_SIZE;
+	else
+		return mal_rx_size(ETH_DATA_LEN + EMAC_MTU_OVERHEAD);
+}
+
+#define EMAC_DMA_ALIGN(x)		ALIGN((x), dma_get_cache_alignment())
+
+#define EMAC_RX_SKB_HEADROOM		\
+	EMAC_DMA_ALIGN(CONFIG_IBM_EMAC_RX_SKB_HEADROOM)
+
+/* Size of RX skb for the given MTU */
+static inline int emac_rx_skb_size(int mtu)
+{
+	int size = max(mtu + EMAC_MTU_OVERHEAD, emac_rx_size(mtu));
+	return EMAC_DMA_ALIGN(size + 2) + EMAC_RX_SKB_HEADROOM;
+}
+
+/* RX DMA sync size */
+static inline int emac_rx_sync_size(int mtu)
+{
+	return EMAC_DMA_ALIGN(emac_rx_size(mtu) + 2);
+}
+
+/* Driver statistcs is split into two parts to make it more cache friendly:
+ *   - normal statistics (packet count, etc)
+ *   - error statistics
+ *
+ * When statistics is requested by ethtool, these parts are concatenated,
+ * normal one goes first.
+ *
+ * Please, keep these structures in sync with emac_stats_keys.
+ */
+
+/* Normal TX/RX Statistics */
+struct emac_stats {
+	u64 rx_packets;
+	u64 rx_bytes;
+	u64 tx_packets;
+	u64 tx_bytes;
+	u64 rx_packets_csum;
+	u64 tx_packets_csum;
+};
+
+/* Error statistics */
+struct emac_error_stats {
+	u64 tx_undo;
+
+	/* Software RX Errors */
+	u64 rx_dropped_stack;
+	u64 rx_dropped_oom;
+	u64 rx_dropped_error;
+	u64 rx_dropped_resize;
+	u64 rx_dropped_mtu;
+	u64 rx_stopped;
+	/* BD reported RX errors */
+	u64 rx_bd_errors;
+	u64 rx_bd_overrun;
+	u64 rx_bd_bad_packet;
+	u64 rx_bd_runt_packet;
+	u64 rx_bd_short_event;
+	u64 rx_bd_alignment_error;
+	u64 rx_bd_bad_fcs;
+	u64 rx_bd_packet_too_long;
+	u64 rx_bd_out_of_range;
+	u64 rx_bd_in_range;
+	/* EMAC IRQ reported RX errors */
+	u64 rx_parity;
+	u64 rx_fifo_overrun;
+	u64 rx_overrun;
+	u64 rx_bad_packet;
+	u64 rx_runt_packet;
+	u64 rx_short_event;
+	u64 rx_alignment_error;
+	u64 rx_bad_fcs;
+	u64 rx_packet_too_long;
+	u64 rx_out_of_range;
+	u64 rx_in_range;
+
+	/* Software TX Errors */
+	u64 tx_dropped;
+	/* BD reported TX errors */
+	u64 tx_bd_errors;
+	u64 tx_bd_bad_fcs;
+	u64 tx_bd_carrier_loss;
+	u64 tx_bd_excessive_deferral;
+	u64 tx_bd_excessive_collisions;
+	u64 tx_bd_late_collision;
+	u64 tx_bd_multple_collisions;
+	u64 tx_bd_single_collision;
+	u64 tx_bd_underrun;
+	u64 tx_bd_sqe;
+	/* EMAC IRQ reported TX errors */
+	u64 tx_parity;
+	u64 tx_underrun;
+	u64 tx_sqe;
+	u64 tx_errors;
+};
+
+#define EMAC_ETHTOOL_STATS_COUNT	((sizeof(struct emac_stats) + \
+					  sizeof(struct emac_error_stats)) \
+					 / sizeof(u64))
+
+struct emac_instance {
+	struct net_device		*ndev;
+	struct resource			rsrc_regs;
+	struct emac_regs		__iomem *emacp;
+	struct platform_device		*ofdev;
+	struct device_node		**blist; /* bootlist entry */
+
+	/* MAL linkage */
+	u32				mal_ph;
+	struct platform_device		*mal_dev;
+	u32				mal_rx_chan;
+	u32				mal_tx_chan;
+	struct mal_instance		*mal;
+	struct mal_commac		commac;
+
+	/* PHY infos */
+	u32				phy_mode;
+	u32				phy_map;
+	u32				phy_address;
+	u32				phy_feat_exc;
+	struct mii_phy			phy;
+	struct mutex			link_lock;
+	struct delayed_work		link_work;
+	int				link_polling;
+
+	/* GPCS PHY infos */
+	u32				gpcs_address;
+
+	/* Shared MDIO if any */
+	u32				mdio_ph;
+	struct platform_device		*mdio_dev;
+	struct emac_instance		*mdio_instance;
+	struct mutex			mdio_lock;
+
+	/* ZMII infos if any */
+	u32				zmii_ph;
+	u32				zmii_port;
+	struct platform_device		*zmii_dev;
+
+	/* RGMII infos if any */
+	u32				rgmii_ph;
+	u32				rgmii_port;
+	struct platform_device		*rgmii_dev;
+
+	/* TAH infos if any */
+	u32				tah_ph;
+	u32				tah_port;
+	struct platform_device		*tah_dev;
+
+	/* IRQs */
+	int				wol_irq;
+	int				emac_irq;
+
+	/* OPB bus frequency in Mhz */
+	u32				opb_bus_freq;
+
+	/* Cell index within an ASIC (for clk mgmnt) */
+	u32				cell_index;
+
+	/* Max supported MTU */
+	u32				max_mtu;
+
+	/* Feature bits (from probe table) */
+	unsigned int			features;
+
+	/* Tx and Rx fifo sizes & other infos in bytes */
+	u32				tx_fifo_size;
+	u32				tx_fifo_size_gige;
+	u32				rx_fifo_size;
+	u32				rx_fifo_size_gige;
+	u32				fifo_entry_size;
+	u32				mal_burst_size; /* move to MAL ? */
+
+	/* IAHT and GAHT filter parameterization */
+	u32				xaht_slots_shift;
+	u32				xaht_width_shift;
+
+	/* Descriptor management
+	 */
+	struct mal_descriptor		*tx_desc;
+	int				tx_cnt;
+	int				tx_slot;
+	int				ack_slot;
+
+	struct mal_descriptor		*rx_desc;
+	int				rx_slot;
+	struct sk_buff			*rx_sg_skb;	/* 1 */
+	int 				rx_skb_size;
+	int				rx_sync_size;
+
+	struct sk_buff			*tx_skb[NUM_TX_BUFF];
+	struct sk_buff			*rx_skb[NUM_RX_BUFF];
+
+	/* Stats
+	 */
+	struct emac_error_stats		estats;
+	struct net_device_stats		nstats;
+	struct emac_stats 		stats;
+
+	/* Misc
+	 */
+	int				reset_failed;
+	int				stop_timeout;	/* in us */
+	int				no_mcast;
+	int				mcast_pending;
+	int				opened;
+	struct work_struct		reset_work;
+	spinlock_t			lock;
+};
+
+/*
+ * Features of various EMAC implementations
+ */
+
+/*
+ * No flow control on 40x according to the original driver
+ */
+#define EMAC_FTR_NO_FLOW_CONTROL_40x	0x00000001
+/*
+ * Cell is an EMAC4
+ */
+#define EMAC_FTR_EMAC4			0x00000002
+/*
+ * For the 440SPe, AMCC inexplicably changed the polarity of
+ * the "operation complete" bit in the MII control register.
+ */
+#define EMAC_FTR_STACR_OC_INVERT	0x00000004
+/*
+ * Set if we have a TAH.
+ */
+#define EMAC_FTR_HAS_TAH		0x00000008
+/*
+ * Set if we have a ZMII.
+ */
+#define EMAC_FTR_HAS_ZMII		0x00000010
+/*
+ * Set if we have a RGMII.
+ */
+#define EMAC_FTR_HAS_RGMII		0x00000020
+/*
+ * Set if we have new type STACR with STAOPC
+ */
+#define EMAC_FTR_HAS_NEW_STACR		0x00000040
+/*
+ * Set if we need phy clock workaround for 440gx
+ */
+#define EMAC_FTR_440GX_PHY_CLK_FIX	0x00000080
+/*
+ * Set if we need phy clock workaround for 440ep or 440gr
+ */
+#define EMAC_FTR_440EP_PHY_CLK_FIX	0x00000100
+/*
+ * The 405EX and 460EX contain the EMAC4SYNC core
+ */
+#define EMAC_FTR_EMAC4SYNC		0x00000200
+/*
+ * Set if we need phy clock workaround for 460ex or 460gt
+ */
+#define EMAC_FTR_460EX_PHY_CLK_FIX	0x00000400
+/*
+ * APM821xx requires Jumbo frame size set explicitly
+ */
+#define EMAC_APM821XX_REQ_JUMBO_FRAME_SIZE	0x00000800
+/*
+ * APM821xx does not support Half Duplex mode
+ */
+#define EMAC_FTR_APM821XX_NO_HALF_DUPLEX	0x00001000
+
+/* Right now, we don't quite handle the always/possible masks on the
+ * most optimal way as we don't have a way to say something like
+ * always EMAC4. Patches welcome.
+ */
+enum {
+	EMAC_FTRS_ALWAYS	= 0,
+
+	EMAC_FTRS_POSSIBLE	=
+#ifdef CONFIG_IBM_EMAC_EMAC4
+	    EMAC_FTR_EMAC4	| EMAC_FTR_EMAC4SYNC	|
+	    EMAC_FTR_HAS_NEW_STACR	|
+	    EMAC_FTR_STACR_OC_INVERT | EMAC_FTR_440GX_PHY_CLK_FIX |
+#endif
+#ifdef CONFIG_IBM_EMAC_TAH
+	    EMAC_FTR_HAS_TAH	|
+#endif
+#ifdef CONFIG_IBM_EMAC_ZMII
+	    EMAC_FTR_HAS_ZMII	|
+#endif
+#ifdef CONFIG_IBM_EMAC_RGMII
+	    EMAC_FTR_HAS_RGMII	|
+#endif
+#ifdef CONFIG_IBM_EMAC_NO_FLOW_CTRL
+	    EMAC_FTR_NO_FLOW_CONTROL_40x |
+#endif
+	EMAC_FTR_460EX_PHY_CLK_FIX |
+	EMAC_FTR_440EP_PHY_CLK_FIX |
+	EMAC_APM821XX_REQ_JUMBO_FRAME_SIZE |
+	EMAC_FTR_APM821XX_NO_HALF_DUPLEX,
+};
+
+static inline int emac_has_feature(struct emac_instance *dev,
+				   unsigned long feature)
+{
+	return (EMAC_FTRS_ALWAYS & feature) ||
+	       (EMAC_FTRS_POSSIBLE & dev->features & feature);
+}
+
+/*
+ * Various instances of the EMAC core have varying 1) number of
+ * address match slots, 2) width of the registers for handling address
+ * match slots, 3) number of registers for handling address match
+ * slots and 4) base offset for those registers.
+ *
+ * These macros and inlines handle these differences based on
+ * parameters supplied by the device structure which are, in turn,
+ * initialized based on the "compatible" entry in the device tree.
+ */
+
+#define	EMAC4_XAHT_SLOTS_SHIFT		6
+#define	EMAC4_XAHT_WIDTH_SHIFT		4
+
+#define	EMAC4SYNC_XAHT_SLOTS_SHIFT	8
+#define	EMAC4SYNC_XAHT_WIDTH_SHIFT	5
+
+#define	EMAC_XAHT_SLOTS(dev)         	(1 << (dev)->xaht_slots_shift)
+#define	EMAC_XAHT_WIDTH(dev)         	(1 << (dev)->xaht_width_shift)
+#define	EMAC_XAHT_REGS(dev)          	(1 << ((dev)->xaht_slots_shift - \
+					       (dev)->xaht_width_shift))
+
+#define	EMAC_XAHT_CRC_TO_SLOT(dev, crc)			\
+	((EMAC_XAHT_SLOTS(dev) - 1) -			\
+	 ((crc) >> ((sizeof (u32) * BITS_PER_BYTE) -	\
+		    (dev)->xaht_slots_shift)))
+
+#define	EMAC_XAHT_SLOT_TO_REG(dev, slot)		\
+	((slot) >> (dev)->xaht_width_shift)
+
+#define	EMAC_XAHT_SLOT_TO_MASK(dev, slot)		\
+	((u32)(1 << (EMAC_XAHT_WIDTH(dev) - 1)) >>	\
+	 ((slot) & (u32)(EMAC_XAHT_WIDTH(dev) - 1)))
+
+static inline u32 *emac_xaht_base(struct emac_instance *dev)
+{
+	struct emac_regs __iomem *p = dev->emacp;
+	int offset;
+
+	/* The first IAHT entry always is the base of the block of
+	 * IAHT and GAHT registers.
+	 */
+	if (emac_has_feature(dev, EMAC_FTR_EMAC4SYNC))
+		offset = offsetof(struct emac_regs, u1.emac4sync.iaht1);
+	else
+		offset = offsetof(struct emac_regs, u0.emac4.iaht1);
+
+	return (u32 *)((ptrdiff_t)p + offset);
+}
+
+static inline u32 *emac_gaht_base(struct emac_instance *dev)
+{
+	/* GAHT registers always come after an identical number of
+	 * IAHT registers.
+	 */
+	return emac_xaht_base(dev) + EMAC_XAHT_REGS(dev);
+}
+
+static inline u32 *emac_iaht_base(struct emac_instance *dev)
+{
+	/* IAHT registers always come before an identical number of
+	 * GAHT registers.
+	 */
+	return emac_xaht_base(dev);
+}
+
+/* Ethtool get_regs complex data.
+ * We want to get not just EMAC registers, but also MAL, ZMII, RGMII, TAH
+ * when available.
+ *
+ * Returned BLOB consists of the ibm_emac_ethtool_regs_hdr,
+ * MAL registers, EMAC registers and optional ZMII, RGMII, TAH registers.
+ * Each register component is preceded with emac_ethtool_regs_subhdr.
+ * Order of the optional headers follows their relative bit posititions
+ * in emac_ethtool_regs_hdr.components
+ */
+#define EMAC_ETHTOOL_REGS_ZMII		0x00000001
+#define EMAC_ETHTOOL_REGS_RGMII		0x00000002
+#define EMAC_ETHTOOL_REGS_TAH		0x00000004
+
+struct emac_ethtool_regs_hdr {
+	u32 components;
+};
+
+struct emac_ethtool_regs_subhdr {
+	u32 version;
+	u32 index;
+};
+
+#define EMAC_ETHTOOL_REGS_VER		0
+#define EMAC4_ETHTOOL_REGS_VER		1
+#define EMAC4SYNC_ETHTOOL_REGS_VER	2
+
+#endif /* __IBM_NEWEMAC_CORE_H */
diff --git a/drivers/net/ethernet/ibm/emac/debug.c b/drivers/net/ethernet/ibm/emac/debug.c
new file mode 100644
index 000000000..a559f326b
--- /dev/null
+++ b/drivers/net/ethernet/ibm/emac/debug.c
@@ -0,0 +1,270 @@
+/*
+ * drivers/net/ethernet/ibm/emac/debug.c
+ *
+ * Driver for PowerPC 4xx on-chip ethernet controller, debug print routines.
+ *
+ * Copyright 2007 Benjamin Herrenschmidt, IBM Corp.
+ *                <benh@kernel.crashing.org>
+ *
+ * Based on the arch/ppc version of the driver:
+ *
+ * Copyright (c) 2004, 2005 Zultys Technologies
+ * Eugene Surovegin <eugene.surovegin@zultys.com> or <ebs@ebshome.net>
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ *
+ */
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/netdevice.h>
+#include <linux/sysrq.h>
+#include <asm/io.h>
+
+#include "core.h"
+
+static DEFINE_SPINLOCK(emac_dbg_lock);
+
+static void emac_desc_dump(struct emac_instance *p)
+{
+	int i;
+	printk("** EMAC %s TX BDs **\n"
+	       " tx_cnt = %d tx_slot = %d ack_slot = %d\n",
+	       p->ofdev->dev.of_node->full_name,
+	       p->tx_cnt, p->tx_slot, p->ack_slot);
+	for (i = 0; i < NUM_TX_BUFF / 2; ++i)
+		printk
+		    ("bd[%2d] 0x%08x %c 0x%04x %4u - bd[%2d] 0x%08x %c 0x%04x %4u\n",
+		     i, p->tx_desc[i].data_ptr, p->tx_skb[i] ? 'V' : ' ',
+		     p->tx_desc[i].ctrl, p->tx_desc[i].data_len,
+		     NUM_TX_BUFF / 2 + i,
+		     p->tx_desc[NUM_TX_BUFF / 2 + i].data_ptr,
+		     p->tx_skb[NUM_TX_BUFF / 2 + i] ? 'V' : ' ',
+		     p->tx_desc[NUM_TX_BUFF / 2 + i].ctrl,
+		     p->tx_desc[NUM_TX_BUFF / 2 + i].data_len);
+
+	printk("** EMAC %s RX BDs **\n"
+	       " rx_slot = %d flags = 0x%lx rx_skb_size = %d rx_sync_size = %d\n"
+	       " rx_sg_skb = 0x%p\n",
+	       p->ofdev->dev.of_node->full_name,
+	       p->rx_slot, p->commac.flags, p->rx_skb_size,
+	       p->rx_sync_size, p->rx_sg_skb);
+	for (i = 0; i < NUM_RX_BUFF / 2; ++i)
+		printk
+		    ("bd[%2d] 0x%08x %c 0x%04x %4u - bd[%2d] 0x%08x %c 0x%04x %4u\n",
+		     i, p->rx_desc[i].data_ptr, p->rx_skb[i] ? 'V' : ' ',
+		     p->rx_desc[i].ctrl, p->rx_desc[i].data_len,
+		     NUM_RX_BUFF / 2 + i,
+		     p->rx_desc[NUM_RX_BUFF / 2 + i].data_ptr,
+		     p->rx_skb[NUM_RX_BUFF / 2 + i] ? 'V' : ' ',
+		     p->rx_desc[NUM_RX_BUFF / 2 + i].ctrl,
+		     p->rx_desc[NUM_RX_BUFF / 2 + i].data_len);
+}
+
+static void emac_mac_dump(struct emac_instance *dev)
+{
+	struct emac_regs __iomem *p = dev->emacp;
+	const int xaht_regs = EMAC_XAHT_REGS(dev);
+	u32 *gaht_base = emac_gaht_base(dev);
+	u32 *iaht_base = emac_iaht_base(dev);
+	int emac4sync = emac_has_feature(dev, EMAC_FTR_EMAC4SYNC);
+	int n;
+
+	printk("** EMAC %s registers **\n"
+	       "MR0 = 0x%08x MR1 = 0x%08x TMR0 = 0x%08x TMR1 = 0x%08x\n"
+	       "RMR = 0x%08x ISR = 0x%08x ISER = 0x%08x\n"
+	       "IAR = %04x%08x VTPID = 0x%04x VTCI = 0x%04x\n",
+	       dev->ofdev->dev.of_node->full_name,
+	       in_be32(&p->mr0), in_be32(&p->mr1),
+	       in_be32(&p->tmr0), in_be32(&p->tmr1),
+	       in_be32(&p->rmr), in_be32(&p->isr), in_be32(&p->iser),
+	       in_be32(&p->iahr), in_be32(&p->ialr), in_be32(&p->vtpid),
+	       in_be32(&p->vtci)
+	       );
+
+	if (emac4sync)
+		printk("MAR = %04x%08x MMAR = %04x%08x\n",
+		       in_be32(&p->u0.emac4sync.mahr),
+		       in_be32(&p->u0.emac4sync.malr),
+		       in_be32(&p->u0.emac4sync.mmahr),
+		       in_be32(&p->u0.emac4sync.mmalr)
+		       );
+
+	for (n = 0; n < xaht_regs; n++)
+		printk("IAHT%02d = 0x%08x\n", n + 1, in_be32(iaht_base + n));
+
+	for (n = 0; n < xaht_regs; n++)
+		printk("GAHT%02d = 0x%08x\n", n + 1, in_be32(gaht_base + n));
+
+	printk("LSA = %04x%08x IPGVR = 0x%04x\n"
+	       "STACR = 0x%08x TRTR = 0x%08x RWMR = 0x%08x\n"
+	       "OCTX = 0x%08x OCRX = 0x%08x\n",
+	       in_be32(&p->lsah), in_be32(&p->lsal), in_be32(&p->ipgvr),
+	       in_be32(&p->stacr), in_be32(&p->trtr), in_be32(&p->rwmr),
+	       in_be32(&p->octx), in_be32(&p->ocrx)
+	       );
+
+	if (!emac4sync) {
+		printk("IPCR = 0x%08x\n",
+		       in_be32(&p->u1.emac4.ipcr)
+		       );
+	} else {
+		printk("REVID = 0x%08x TPC = 0x%08x\n",
+		       in_be32(&p->u1.emac4sync.revid),
+		       in_be32(&p->u1.emac4sync.tpc)
+		       );
+	}
+
+	emac_desc_dump(dev);
+}
+
+static void emac_mal_dump(struct mal_instance *mal)
+{
+	int i;
+
+	printk("** MAL %s Registers **\n"
+	       "CFG = 0x%08x ESR = 0x%08x IER = 0x%08x\n"
+	       "TX|CASR = 0x%08x CARR = 0x%08x EOBISR = 0x%08x DEIR = 0x%08x\n"
+	       "RX|CASR = 0x%08x CARR = 0x%08x EOBISR = 0x%08x DEIR = 0x%08x\n",
+	       mal->ofdev->dev.of_node->full_name,
+	       get_mal_dcrn(mal, MAL_CFG), get_mal_dcrn(mal, MAL_ESR),
+	       get_mal_dcrn(mal, MAL_IER),
+	       get_mal_dcrn(mal, MAL_TXCASR), get_mal_dcrn(mal, MAL_TXCARR),
+	       get_mal_dcrn(mal, MAL_TXEOBISR), get_mal_dcrn(mal, MAL_TXDEIR),
+	       get_mal_dcrn(mal, MAL_RXCASR), get_mal_dcrn(mal, MAL_RXCARR),
+	       get_mal_dcrn(mal, MAL_RXEOBISR), get_mal_dcrn(mal, MAL_RXDEIR)
+	    );
+
+	printk("TX|");
+	for (i = 0; i < mal->num_tx_chans; ++i) {
+		if (i && !(i % 4))
+			printk("\n   ");
+		printk("CTP%d = 0x%08x ", i, get_mal_dcrn(mal, MAL_TXCTPR(i)));
+	}
+	printk("\nRX|");
+	for (i = 0; i < mal->num_rx_chans; ++i) {
+		if (i && !(i % 4))
+			printk("\n   ");
+		printk("CTP%d = 0x%08x ", i, get_mal_dcrn(mal, MAL_RXCTPR(i)));
+	}
+	printk("\n   ");
+	for (i = 0; i < mal->num_rx_chans; ++i) {
+		u32 r = get_mal_dcrn(mal, MAL_RCBS(i));
+		if (i && !(i % 3))
+			printk("\n   ");
+		printk("RCBS%d = 0x%08x (%d) ", i, r, r * 16);
+	}
+	printk("\n");
+}
+
+static struct emac_instance *__emacs[4];
+static struct mal_instance *__mals[1];
+
+void emac_dbg_register(struct emac_instance *dev)
+{
+	unsigned long flags;
+	int i;
+
+	spin_lock_irqsave(&emac_dbg_lock, flags);
+	for (i = 0; i < ARRAY_SIZE(__emacs); i++)
+		if (__emacs[i] == NULL) {
+			__emacs[i] = dev;
+			break;
+		}
+	spin_unlock_irqrestore(&emac_dbg_lock, flags);
+}
+
+void emac_dbg_unregister(struct emac_instance *dev)
+{
+	unsigned long flags;
+	int i;
+
+	spin_lock_irqsave(&emac_dbg_lock, flags);
+	for (i = 0; i < ARRAY_SIZE(__emacs); i++)
+		if (__emacs[i] == dev) {
+			__emacs[i] = NULL;
+			break;
+		}
+	spin_unlock_irqrestore(&emac_dbg_lock, flags);
+}
+
+void mal_dbg_register(struct mal_instance *mal)
+{
+	unsigned long flags;
+	int i;
+
+	spin_lock_irqsave(&emac_dbg_lock, flags);
+	for (i = 0; i < ARRAY_SIZE(__mals); i++)
+		if (__mals[i] == NULL) {
+			__mals[i] = mal;
+			break;
+		}
+	spin_unlock_irqrestore(&emac_dbg_lock, flags);
+}
+
+void mal_dbg_unregister(struct mal_instance *mal)
+{
+	unsigned long flags;
+	int i;
+
+	spin_lock_irqsave(&emac_dbg_lock, flags);
+	for (i = 0; i < ARRAY_SIZE(__mals); i++)
+		if (__mals[i] == mal) {
+			__mals[i] = NULL;
+			break;
+		}
+	spin_unlock_irqrestore(&emac_dbg_lock, flags);
+}
+
+void emac_dbg_dump_all(void)
+{
+	unsigned int i;
+	unsigned long flags;
+
+	spin_lock_irqsave(&emac_dbg_lock, flags);
+
+	for (i = 0; i < ARRAY_SIZE(__mals); ++i)
+		if (__mals[i])
+			emac_mal_dump(__mals[i]);
+
+	for (i = 0; i < ARRAY_SIZE(__emacs); ++i)
+		if (__emacs[i])
+			emac_mac_dump(__emacs[i]);
+
+	spin_unlock_irqrestore(&emac_dbg_lock, flags);
+}
+
+#if defined(CONFIG_MAGIC_SYSRQ)
+static void emac_sysrq_handler(int key)
+{
+	emac_dbg_dump_all();
+}
+
+static struct sysrq_key_op emac_sysrq_op = {
+	.handler = emac_sysrq_handler,
+	.help_msg = "emac(c)",
+	.action_msg = "Show EMAC(s) status",
+};
+
+int __init emac_init_debug(void)
+{
+	return register_sysrq_key('c', &emac_sysrq_op);
+}
+
+void __exit emac_fini_debug(void)
+{
+	unregister_sysrq_key('c', &emac_sysrq_op);
+}
+
+#else
+int __init emac_init_debug(void)
+{
+	return 0;
+}
+void __exit emac_fini_debug(void)
+{
+}
+#endif				/* CONFIG_MAGIC_SYSRQ */
diff --git a/drivers/net/ethernet/ibm/emac/debug.h b/drivers/net/ethernet/ibm/emac/debug.h
new file mode 100644
index 000000000..9c45efe4c
--- /dev/null
+++ b/drivers/net/ethernet/ibm/emac/debug.h
@@ -0,0 +1,83 @@
+/*
+ * drivers/net/ethernet/ibm/emac/debug.h
+ *
+ * Driver for PowerPC 4xx on-chip ethernet controller, debug print routines.
+ *
+ * Copyright 2007 Benjamin Herrenschmidt, IBM Corp.
+ *                <benh@kernel.crashing.org>
+ *
+ * Based on the arch/ppc version of the driver:
+ *
+ * Copyright (c) 2004, 2005 Zultys Technologies
+ * Eugene Surovegin <eugene.surovegin@zultys.com> or <ebs@ebshome.net>
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ *
+ */
+#ifndef __IBM_NEWEMAC_DEBUG_H
+#define __IBM_NEWEMAC_DEBUG_H
+
+#include <linux/init.h>
+
+#include "core.h"
+
+#if defined(CONFIG_IBM_EMAC_DEBUG)
+
+struct emac_instance;
+struct mal_instance;
+
+void emac_dbg_register(struct emac_instance *dev);
+void emac_dbg_unregister(struct emac_instance *dev);
+void mal_dbg_register(struct mal_instance *mal);
+void mal_dbg_unregister(struct mal_instance *mal);
+int emac_init_debug(void) __init;
+void emac_fini_debug(void) __exit;
+void emac_dbg_dump_all(void);
+
+# define DBG_LEVEL		1
+
+#else
+
+# define emac_dbg_register(x)	do { } while(0)
+# define emac_dbg_unregister(x)	do { } while(0)
+# define mal_dbg_register(x)	do { } while(0)
+# define mal_dbg_unregister(x)	do { } while(0)
+# define emac_init_debug()	do { } while(0)
+# define emac_fini_debug()	do { } while(0)
+# define emac_dbg_dump_all()	do { } while(0)
+
+# define DBG_LEVEL		0
+
+#endif
+
+#define EMAC_DBG(d, name, fmt, arg...) \
+	printk(KERN_DEBUG #name "%s: " fmt, d->ofdev->dev.of_node->full_name, ## arg)
+
+#if DBG_LEVEL > 0
+#  define DBG(d,f,x...)		EMAC_DBG(d, emac, f, ##x)
+#  define MAL_DBG(d,f,x...)	EMAC_DBG(d, mal, f, ##x)
+#  define ZMII_DBG(d,f,x...)	EMAC_DBG(d, zmii, f, ##x)
+#  define RGMII_DBG(d,f,x...)	EMAC_DBG(d, rgmii, f, ##x)
+#  define NL			"\n"
+#else
+#  define DBG(f,x...)		((void)0)
+#  define MAL_DBG(d,f,x...)	((void)0)
+#  define ZMII_DBG(d,f,x...)	((void)0)
+#  define RGMII_DBG(d,f,x...)	((void)0)
+#endif
+#if DBG_LEVEL > 1
+#  define DBG2(d,f,x...) 	DBG(d,f, ##x)
+#  define MAL_DBG2(d,f,x...) 	MAL_DBG(d,f, ##x)
+#  define ZMII_DBG2(d,f,x...) 	ZMII_DBG(d,f, ##x)
+#  define RGMII_DBG2(d,f,x...) 	RGMII_DBG(d,f, ##x)
+#else
+#  define DBG2(f,x...) 		((void)0)
+#  define MAL_DBG2(d,f,x...) 	((void)0)
+#  define ZMII_DBG2(d,f,x...) 	((void)0)
+#  define RGMII_DBG2(d,f,x...) 	((void)0)
+#endif
+
+#endif /* __IBM_NEWEMAC_DEBUG_H */
diff --git a/drivers/net/ethernet/ibm/emac/emac.h b/drivers/net/ethernet/ibm/emac/emac.h
new file mode 100644
index 000000000..5afcc27ce
--- /dev/null
+++ b/drivers/net/ethernet/ibm/emac/emac.h
@@ -0,0 +1,314 @@
+/*
+ * drivers/net/ethernet/ibm/emac/emac.h
+ *
+ * Register definitions for PowerPC 4xx on-chip ethernet contoller
+ *
+ * Copyright 2007 Benjamin Herrenschmidt, IBM Corp.
+ *                <benh@kernel.crashing.org>
+ *
+ * Based on the arch/ppc version of the driver:
+ *
+ * Copyright (c) 2004, 2005 Zultys Technologies.
+ * Eugene Surovegin <eugene.surovegin@zultys.com> or <ebs@ebshome.net>
+ *
+ * Based on original work by
+ *      Matt Porter <mporter@kernel.crashing.org>
+ *      Armin Kuster <akuster@mvista.com>
+ * 	Copyright 2002-2004 MontaVista Software Inc.
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ *
+ */
+#ifndef __IBM_NEWEMAC_H
+#define __IBM_NEWEMAC_H
+
+#include <linux/types.h>
+#include <linux/phy.h>
+
+/* EMAC registers 			Write Access rules */
+struct emac_regs {
+	/* Common registers across all EMAC implementations. */
+	u32 mr0;			/* Special 	*/
+	u32 mr1;			/* Reset 	*/
+	u32 tmr0;			/* Special 	*/
+	u32 tmr1;			/* Special 	*/
+	u32 rmr;			/* Reset 	*/
+	u32 isr;			/* Always 	*/
+	u32 iser;			/* Reset 	*/
+	u32 iahr;			/* Reset, R, T 	*/
+	u32 ialr;			/* Reset, R, T 	*/
+	u32 vtpid;			/* Reset, R, T 	*/
+	u32 vtci;			/* Reset, R, T 	*/
+	u32 ptr;			/* Reset,    T 	*/
+	union {
+		/* Registers unique to EMAC4 implementations */
+		struct {
+			u32 iaht1;	/* Reset, R	*/
+			u32 iaht2;	/* Reset, R	*/
+			u32 iaht3;	/* Reset, R	*/
+			u32 iaht4;	/* Reset, R	*/
+			u32 gaht1;	/* Reset, R	*/
+			u32 gaht2;	/* Reset, R	*/
+			u32 gaht3;	/* Reset, R	*/
+			u32 gaht4;	/* Reset, R	*/
+		} emac4;
+		/* Registers unique to EMAC4SYNC implementations */
+		struct {
+			u32 mahr;	/* Reset, R, T  */
+			u32 malr;	/* Reset, R, T  */
+			u32 mmahr;	/* Reset, R, T  */
+			u32 mmalr;	/* Reset, R, T  */
+			u32 rsvd0[4];
+		} emac4sync;
+	} u0;
+	/* Common registers across all EMAC implementations. */
+	u32 lsah;
+	u32 lsal;
+	u32 ipgvr;			/* Reset,    T 	*/
+	u32 stacr;			/* Special 	*/
+	u32 trtr;			/* Special 	*/
+	u32 rwmr;			/* Reset 	*/
+	u32 octx;
+	u32 ocrx;
+	union {
+		/* Registers unique to EMAC4 implementations */
+		struct {
+			u32 ipcr;
+		} emac4;
+		/* Registers unique to EMAC4SYNC implementations */
+		struct {
+			u32 rsvd1;
+			u32 revid;
+ 			u32 rsvd2[2];
+			u32 iaht1;	/* Reset, R     */
+			u32 iaht2;	/* Reset, R     */
+			u32 iaht3;	/* Reset, R     */
+			u32 iaht4;	/* Reset, R     */
+			u32 iaht5;	/* Reset, R     */
+			u32 iaht6;	/* Reset, R     */
+			u32 iaht7;	/* Reset, R     */
+			u32 iaht8;	/* Reset, R     */
+			u32 gaht1;	/* Reset, R     */
+			u32 gaht2;	/* Reset, R     */
+			u32 gaht3;	/* Reset, R     */
+			u32 gaht4;	/* Reset, R     */
+			u32 gaht5;	/* Reset, R     */
+			u32 gaht6;	/* Reset, R     */
+			u32 gaht7;	/* Reset, R     */
+			u32 gaht8;	/* Reset, R     */
+			u32 tpc;	/* Reset, T     */
+		} emac4sync;
+	} u1;
+};
+
+/*
+ * PHY mode settings (EMAC <-> ZMII/RGMII bridge <-> PHY)
+ */
+#define PHY_MODE_NA	PHY_INTERFACE_MODE_NA
+#define PHY_MODE_MII	PHY_INTERFACE_MODE_MII
+#define PHY_MODE_RMII	PHY_INTERFACE_MODE_RMII
+#define PHY_MODE_SMII	PHY_INTERFACE_MODE_SMII
+#define PHY_MODE_RGMII	PHY_INTERFACE_MODE_RGMII
+#define PHY_MODE_TBI	PHY_INTERFACE_MODE_TBI
+#define PHY_MODE_GMII	PHY_INTERFACE_MODE_GMII
+#define PHY_MODE_RTBI	PHY_INTERFACE_MODE_RTBI
+#define PHY_MODE_SGMII	PHY_INTERFACE_MODE_SGMII
+
+/* EMACx_MR0 */
+#define EMAC_MR0_RXI			0x80000000
+#define EMAC_MR0_TXI			0x40000000
+#define EMAC_MR0_SRST			0x20000000
+#define EMAC_MR0_TXE			0x10000000
+#define EMAC_MR0_RXE			0x08000000
+#define EMAC_MR0_WKE			0x04000000
+
+/* EMACx_MR1 */
+#define EMAC_MR1_FDE			0x80000000
+#define EMAC_MR1_ILE			0x40000000
+#define EMAC_MR1_VLE			0x20000000
+#define EMAC_MR1_EIFC			0x10000000
+#define EMAC_MR1_APP			0x08000000
+#define EMAC_MR1_IST			0x01000000
+
+#define EMAC_MR1_MF_MASK		0x00c00000
+#define EMAC_MR1_MF_10			0x00000000
+#define EMAC_MR1_MF_100			0x00400000
+#define EMAC_MR1_MF_1000		0x00800000
+#define EMAC_MR1_MF_1000GPCS		0x00c00000
+#define EMAC_MR1_MF_IPPA(id)		(((id) & 0x1f) << 6)
+
+#define EMAC_MR1_RFS_4K			0x00300000
+#define EMAC_MR1_RFS_16K		0x00000000
+#define EMAC_MR1_TFS_2K			0x00080000
+#define EMAC_MR1_TR0_MULT		0x00008000
+#define EMAC_MR1_JPSM			0x00000000
+#define EMAC_MR1_MWSW_001		0x00000000
+#define EMAC_MR1_BASE(opb)		(EMAC_MR1_TFS_2K | EMAC_MR1_TR0_MULT)
+
+
+#define EMAC4_MR1_RFS_2K		0x00100000
+#define EMAC4_MR1_RFS_4K		0x00180000
+#define EMAC4_MR1_RFS_16K		0x00280000
+#define EMAC4_MR1_TFS_2K       		0x00020000
+#define EMAC4_MR1_TFS_4K		0x00030000
+#define EMAC4_MR1_TFS_16K		0x00050000
+#define EMAC4_MR1_TR			0x00008000
+#define EMAC4_MR1_MWSW_001		0x00001000
+#define EMAC4_MR1_JPSM			0x00000800
+#define EMAC4_MR1_OBCI_MASK		0x00000038
+#define EMAC4_MR1_OBCI_50		0x00000000
+#define EMAC4_MR1_OBCI_66		0x00000008
+#define EMAC4_MR1_OBCI_83		0x00000010
+#define EMAC4_MR1_OBCI_100		0x00000018
+#define EMAC4_MR1_OBCI_100P		0x00000020
+#define EMAC4_MR1_OBCI(freq)		((freq) <= 50  ? EMAC4_MR1_OBCI_50 : \
+					 (freq) <= 66  ? EMAC4_MR1_OBCI_66 : \
+					 (freq) <= 83  ? EMAC4_MR1_OBCI_83 : \
+					 (freq) <= 100 ? EMAC4_MR1_OBCI_100 : \
+						EMAC4_MR1_OBCI_100P)
+
+/* EMACx_TMR0 */
+#define EMAC_TMR0_GNP			0x80000000
+#define EMAC_TMR0_DEFAULT		0x00000000
+#define EMAC4_TMR0_TFAE_2_32		0x00000001
+#define EMAC4_TMR0_TFAE_4_64		0x00000002
+#define EMAC4_TMR0_TFAE_8_128		0x00000003
+#define EMAC4_TMR0_TFAE_16_256		0x00000004
+#define EMAC4_TMR0_TFAE_32_512		0x00000005
+#define EMAC4_TMR0_TFAE_64_1024		0x00000006
+#define EMAC4_TMR0_TFAE_128_2048	0x00000007
+#define EMAC4_TMR0_DEFAULT		EMAC4_TMR0_TFAE_2_32
+#define EMAC_TMR0_XMIT			(EMAC_TMR0_GNP | EMAC_TMR0_DEFAULT)
+#define EMAC4_TMR0_XMIT			(EMAC_TMR0_GNP | EMAC4_TMR0_DEFAULT)
+
+/* EMACx_TMR1 */
+
+#define EMAC_TMR1(l,h)			(((l) << 27) | (((h) & 0xff) << 16))
+#define EMAC4_TMR1(l,h)			(((l) << 27) | (((h) & 0x3ff) << 14))
+
+/* EMACx_RMR */
+#define EMAC_RMR_SP			0x80000000
+#define EMAC_RMR_SFCS			0x40000000
+#define EMAC_RMR_RRP			0x20000000
+#define EMAC_RMR_RFP			0x10000000
+#define EMAC_RMR_ROP			0x08000000
+#define EMAC_RMR_RPIR			0x04000000
+#define EMAC_RMR_PPP			0x02000000
+#define EMAC_RMR_PME			0x01000000
+#define EMAC_RMR_PMME			0x00800000
+#define EMAC_RMR_IAE			0x00400000
+#define EMAC_RMR_MIAE			0x00200000
+#define EMAC_RMR_BAE			0x00100000
+#define EMAC_RMR_MAE			0x00080000
+#define EMAC_RMR_BASE			0x00000000
+#define EMAC4_RMR_RFAF_2_32		0x00000001
+#define EMAC4_RMR_RFAF_4_64		0x00000002
+#define EMAC4_RMR_RFAF_8_128		0x00000003
+#define EMAC4_RMR_RFAF_16_256		0x00000004
+#define EMAC4_RMR_RFAF_32_512		0x00000005
+#define EMAC4_RMR_RFAF_64_1024		0x00000006
+#define EMAC4_RMR_RFAF_128_2048		0x00000007
+#define EMAC4_RMR_BASE			EMAC4_RMR_RFAF_128_2048
+#define EMAC4_RMR_MJS_MASK              0x0001fff8
+#define EMAC4_RMR_MJS(s)                (((s) << 3) & EMAC4_RMR_MJS_MASK)
+
+/* EMACx_ISR & EMACx_ISER */
+#define EMAC4_ISR_TXPE			0x20000000
+#define EMAC4_ISR_RXPE			0x10000000
+#define EMAC4_ISR_TXUE			0x08000000
+#define EMAC4_ISR_RXOE			0x04000000
+#define EMAC_ISR_OVR			0x02000000
+#define EMAC_ISR_PP			0x01000000
+#define EMAC_ISR_BP			0x00800000
+#define EMAC_ISR_RP			0x00400000
+#define EMAC_ISR_SE			0x00200000
+#define EMAC_ISR_ALE			0x00100000
+#define EMAC_ISR_BFCS			0x00080000
+#define EMAC_ISR_PTLE			0x00040000
+#define EMAC_ISR_ORE			0x00020000
+#define EMAC_ISR_IRE			0x00010000
+#define EMAC_ISR_SQE			0x00000080
+#define EMAC_ISR_TE			0x00000040
+#define EMAC_ISR_MOS			0x00000002
+#define EMAC_ISR_MOF			0x00000001
+
+/* EMACx_STACR */
+#define EMAC_STACR_PHYD_MASK		0xffff
+#define EMAC_STACR_PHYD_SHIFT		16
+#define EMAC_STACR_OC			0x00008000
+#define EMAC_STACR_PHYE			0x00004000
+#define EMAC_STACR_STAC_MASK		0x00003000
+#define EMAC_STACR_STAC_READ		0x00001000
+#define EMAC_STACR_STAC_WRITE		0x00002000
+#define EMAC_STACR_OPBC_MASK		0x00000C00
+#define EMAC_STACR_OPBC_50		0x00000000
+#define EMAC_STACR_OPBC_66		0x00000400
+#define EMAC_STACR_OPBC_83		0x00000800
+#define EMAC_STACR_OPBC_100		0x00000C00
+#define EMAC_STACR_OPBC(freq)		((freq) <= 50 ? EMAC_STACR_OPBC_50 : \
+					 (freq) <= 66 ? EMAC_STACR_OPBC_66 : \
+					 (freq) <= 83 ? EMAC_STACR_OPBC_83 : EMAC_STACR_OPBC_100)
+#define EMAC_STACR_BASE(opb)		EMAC_STACR_OPBC(opb)
+#define EMAC4_STACR_BASE(opb)		0x00000000
+#define EMAC_STACR_PCDA_MASK		0x1f
+#define EMAC_STACR_PCDA_SHIFT		5
+#define EMAC_STACR_PRA_MASK		0x1f
+#define EMACX_STACR_STAC_MASK		0x00003800
+#define EMACX_STACR_STAC_READ		0x00001000
+#define EMACX_STACR_STAC_WRITE		0x00000800
+#define EMACX_STACR_STAC_IND_ADDR	0x00002000
+#define EMACX_STACR_STAC_IND_READ	0x00003800
+#define EMACX_STACR_STAC_IND_READINC	0x00003000
+#define EMACX_STACR_STAC_IND_WRITE	0x00002800
+
+
+/* EMACx_TRTR */
+#define EMAC_TRTR_SHIFT_EMAC4		24
+#define EMAC_TRTR_SHIFT		27
+
+/* EMAC specific TX descriptor control fields (write access) */
+#define EMAC_TX_CTRL_GFCS		0x0200
+#define EMAC_TX_CTRL_GP			0x0100
+#define EMAC_TX_CTRL_ISA		0x0080
+#define EMAC_TX_CTRL_RSA		0x0040
+#define EMAC_TX_CTRL_IVT		0x0020
+#define EMAC_TX_CTRL_RVT		0x0010
+#define EMAC_TX_CTRL_TAH_CSUM		0x000e
+
+/* EMAC specific TX descriptor status fields (read access) */
+#define EMAC_TX_ST_BFCS			0x0200
+#define EMAC_TX_ST_LCS			0x0080
+#define EMAC_TX_ST_ED			0x0040
+#define EMAC_TX_ST_EC			0x0020
+#define EMAC_TX_ST_LC			0x0010
+#define EMAC_TX_ST_MC			0x0008
+#define EMAC_TX_ST_SC			0x0004
+#define EMAC_TX_ST_UR			0x0002
+#define EMAC_TX_ST_SQE			0x0001
+#define EMAC_IS_BAD_TX			(EMAC_TX_ST_LCS | EMAC_TX_ST_ED | \
+					 EMAC_TX_ST_EC | EMAC_TX_ST_LC | \
+					 EMAC_TX_ST_MC | EMAC_TX_ST_UR)
+#define EMAC_IS_BAD_TX_TAH		(EMAC_TX_ST_LCS | EMAC_TX_ST_ED | \
+					 EMAC_TX_ST_EC | EMAC_TX_ST_LC)
+
+/* EMAC specific RX descriptor status fields (read access) */
+#define EMAC_RX_ST_OE			0x0200
+#define EMAC_RX_ST_PP			0x0100
+#define EMAC_RX_ST_BP			0x0080
+#define EMAC_RX_ST_RP			0x0040
+#define EMAC_RX_ST_SE			0x0020
+#define EMAC_RX_ST_AE			0x0010
+#define EMAC_RX_ST_BFCS			0x0008
+#define EMAC_RX_ST_PTL			0x0004
+#define EMAC_RX_ST_ORE			0x0002
+#define EMAC_RX_ST_IRE			0x0001
+#define EMAC_RX_TAH_BAD_CSUM		0x0003
+#define EMAC_BAD_RX_MASK		(EMAC_RX_ST_OE | EMAC_RX_ST_BP | \
+					 EMAC_RX_ST_RP | EMAC_RX_ST_SE | \
+					 EMAC_RX_ST_AE | EMAC_RX_ST_BFCS | \
+					 EMAC_RX_ST_PTL | EMAC_RX_ST_ORE | \
+					 EMAC_RX_ST_IRE )
+#endif /* __IBM_NEWEMAC_H */
diff --git a/drivers/net/ethernet/ibm/emac/mal.c b/drivers/net/ethernet/ibm/emac/mal.c
new file mode 100644
index 000000000..fdb5cdb3c
--- /dev/null
+++ b/drivers/net/ethernet/ibm/emac/mal.c
@@ -0,0 +1,793 @@
+/*
+ * drivers/net/ethernet/ibm/emac/mal.c
+ *
+ * Memory Access Layer (MAL) support
+ *
+ * Copyright 2007 Benjamin Herrenschmidt, IBM Corp.
+ *                <benh@kernel.crashing.org>
+ *
+ * Based on the arch/ppc version of the driver:
+ *
+ * Copyright (c) 2004, 2005 Zultys Technologies.
+ * Eugene Surovegin <eugene.surovegin@zultys.com> or <ebs@ebshome.net>
+ *
+ * Based on original work by
+ *      Benjamin Herrenschmidt <benh@kernel.crashing.org>,
+ *      David Gibson <hermes@gibson.dropbear.id.au>,
+ *
+ *      Armin Kuster <akuster@mvista.com>
+ *      Copyright 2002 MontaVista Softare Inc.
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ *
+ */
+
+#include <linux/delay.h>
+#include <linux/slab.h>
+#include <linux/of_irq.h>
+
+#include "core.h"
+#include <asm/dcr-regs.h>
+
+static int mal_count;
+
+int mal_register_commac(struct mal_instance *mal, struct mal_commac *commac)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&mal->lock, flags);
+
+	MAL_DBG(mal, "reg(%08x, %08x)" NL,
+		commac->tx_chan_mask, commac->rx_chan_mask);
+
+	/* Don't let multiple commacs claim the same channel(s) */
+	if ((mal->tx_chan_mask & commac->tx_chan_mask) ||
+	    (mal->rx_chan_mask & commac->rx_chan_mask)) {
+		spin_unlock_irqrestore(&mal->lock, flags);
+		printk(KERN_WARNING "mal%d: COMMAC channels conflict!\n",
+		       mal->index);
+		return -EBUSY;
+	}
+
+	if (list_empty(&mal->list))
+		napi_enable(&mal->napi);
+	mal->tx_chan_mask |= commac->tx_chan_mask;
+	mal->rx_chan_mask |= commac->rx_chan_mask;
+	list_add(&commac->list, &mal->list);
+
+	spin_unlock_irqrestore(&mal->lock, flags);
+
+	return 0;
+}
+
+void mal_unregister_commac(struct mal_instance	*mal,
+		struct mal_commac *commac)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&mal->lock, flags);
+
+	MAL_DBG(mal, "unreg(%08x, %08x)" NL,
+		commac->tx_chan_mask, commac->rx_chan_mask);
+
+	mal->tx_chan_mask &= ~commac->tx_chan_mask;
+	mal->rx_chan_mask &= ~commac->rx_chan_mask;
+	list_del_init(&commac->list);
+	if (list_empty(&mal->list))
+		napi_disable(&mal->napi);
+
+	spin_unlock_irqrestore(&mal->lock, flags);
+}
+
+int mal_set_rcbs(struct mal_instance *mal, int channel, unsigned long size)
+{
+	BUG_ON(channel < 0 || channel >= mal->num_rx_chans ||
+	       size > MAL_MAX_RX_SIZE);
+
+	MAL_DBG(mal, "set_rbcs(%d, %lu)" NL, channel, size);
+
+	if (size & 0xf) {
+		printk(KERN_WARNING
+		       "mal%d: incorrect RX size %lu for the channel %d\n",
+		       mal->index, size, channel);
+		return -EINVAL;
+	}
+
+	set_mal_dcrn(mal, MAL_RCBS(channel), size >> 4);
+	return 0;
+}
+
+int mal_tx_bd_offset(struct mal_instance *mal, int channel)
+{
+	BUG_ON(channel < 0 || channel >= mal->num_tx_chans);
+
+	return channel * NUM_TX_BUFF;
+}
+
+int mal_rx_bd_offset(struct mal_instance *mal, int channel)
+{
+	BUG_ON(channel < 0 || channel >= mal->num_rx_chans);
+	return mal->num_tx_chans * NUM_TX_BUFF + channel * NUM_RX_BUFF;
+}
+
+void mal_enable_tx_channel(struct mal_instance *mal, int channel)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&mal->lock, flags);
+
+	MAL_DBG(mal, "enable_tx(%d)" NL, channel);
+
+	set_mal_dcrn(mal, MAL_TXCASR,
+		     get_mal_dcrn(mal, MAL_TXCASR) | MAL_CHAN_MASK(channel));
+
+	spin_unlock_irqrestore(&mal->lock, flags);
+}
+
+void mal_disable_tx_channel(struct mal_instance *mal, int channel)
+{
+	set_mal_dcrn(mal, MAL_TXCARR, MAL_CHAN_MASK(channel));
+
+	MAL_DBG(mal, "disable_tx(%d)" NL, channel);
+}
+
+void mal_enable_rx_channel(struct mal_instance *mal, int channel)
+{
+	unsigned long flags;
+
+	/*
+	 * On some 4xx PPC's (e.g. 460EX/GT), the rx channel is a multiple
+	 * of 8, but enabling in MAL_RXCASR needs the divided by 8 value
+	 * for the bitmask
+	 */
+	if (!(channel % 8))
+		channel >>= 3;
+
+	spin_lock_irqsave(&mal->lock, flags);
+
+	MAL_DBG(mal, "enable_rx(%d)" NL, channel);
+
+	set_mal_dcrn(mal, MAL_RXCASR,
+		     get_mal_dcrn(mal, MAL_RXCASR) | MAL_CHAN_MASK(channel));
+
+	spin_unlock_irqrestore(&mal->lock, flags);
+}
+
+void mal_disable_rx_channel(struct mal_instance *mal, int channel)
+{
+	/*
+	 * On some 4xx PPC's (e.g. 460EX/GT), the rx channel is a multiple
+	 * of 8, but enabling in MAL_RXCASR needs the divided by 8 value
+	 * for the bitmask
+	 */
+	if (!(channel % 8))
+		channel >>= 3;
+
+	set_mal_dcrn(mal, MAL_RXCARR, MAL_CHAN_MASK(channel));
+
+	MAL_DBG(mal, "disable_rx(%d)" NL, channel);
+}
+
+void mal_poll_add(struct mal_instance *mal, struct mal_commac *commac)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&mal->lock, flags);
+
+	MAL_DBG(mal, "poll_add(%p)" NL, commac);
+
+	/* starts disabled */
+	set_bit(MAL_COMMAC_POLL_DISABLED, &commac->flags);
+
+	list_add_tail(&commac->poll_list, &mal->poll_list);
+
+	spin_unlock_irqrestore(&mal->lock, flags);
+}
+
+void mal_poll_del(struct mal_instance *mal, struct mal_commac *commac)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&mal->lock, flags);
+
+	MAL_DBG(mal, "poll_del(%p)" NL, commac);
+
+	list_del(&commac->poll_list);
+
+	spin_unlock_irqrestore(&mal->lock, flags);
+}
+
+/* synchronized by mal_poll() */
+static inline void mal_enable_eob_irq(struct mal_instance *mal)
+{
+	MAL_DBG2(mal, "enable_irq" NL);
+
+	// XXX might want to cache MAL_CFG as the DCR read can be slooooow
+	set_mal_dcrn(mal, MAL_CFG, get_mal_dcrn(mal, MAL_CFG) | MAL_CFG_EOPIE);
+}
+
+/* synchronized by NAPI state */
+static inline void mal_disable_eob_irq(struct mal_instance *mal)
+{
+	// XXX might want to cache MAL_CFG as the DCR read can be slooooow
+	set_mal_dcrn(mal, MAL_CFG, get_mal_dcrn(mal, MAL_CFG) & ~MAL_CFG_EOPIE);
+
+	MAL_DBG2(mal, "disable_irq" NL);
+}
+
+static irqreturn_t mal_serr(int irq, void *dev_instance)
+{
+	struct mal_instance *mal = dev_instance;
+
+	u32 esr = get_mal_dcrn(mal, MAL_ESR);
+
+	/* Clear the error status register */
+	set_mal_dcrn(mal, MAL_ESR, esr);
+
+	MAL_DBG(mal, "SERR %08x" NL, esr);
+
+	if (esr & MAL_ESR_EVB) {
+		if (esr & MAL_ESR_DE) {
+			/* We ignore Descriptor error,
+			 * TXDE or RXDE interrupt will be generated anyway.
+			 */
+			return IRQ_HANDLED;
+		}
+
+		if (esr & MAL_ESR_PEIN) {
+			/* PLB error, it's probably buggy hardware or
+			 * incorrect physical address in BD (i.e. bug)
+			 */
+			if (net_ratelimit())
+				printk(KERN_ERR
+				       "mal%d: system error, "
+				       "PLB (ESR = 0x%08x)\n",
+				       mal->index, esr);
+			return IRQ_HANDLED;
+		}
+
+		/* OPB error, it's probably buggy hardware or incorrect
+		 * EBC setup
+		 */
+		if (net_ratelimit())
+			printk(KERN_ERR
+			       "mal%d: system error, OPB (ESR = 0x%08x)\n",
+			       mal->index, esr);
+	}
+	return IRQ_HANDLED;
+}
+
+static inline void mal_schedule_poll(struct mal_instance *mal)
+{
+	if (likely(napi_schedule_prep(&mal->napi))) {
+		MAL_DBG2(mal, "schedule_poll" NL);
+		spin_lock(&mal->lock);
+		mal_disable_eob_irq(mal);
+		spin_unlock(&mal->lock);
+		__napi_schedule(&mal->napi);
+	} else
+		MAL_DBG2(mal, "already in poll" NL);
+}
+
+static irqreturn_t mal_txeob(int irq, void *dev_instance)
+{
+	struct mal_instance *mal = dev_instance;
+
+	u32 r = get_mal_dcrn(mal, MAL_TXEOBISR);
+
+	MAL_DBG2(mal, "txeob %08x" NL, r);
+
+	mal_schedule_poll(mal);
+	set_mal_dcrn(mal, MAL_TXEOBISR, r);
+
+#ifdef CONFIG_PPC_DCR_NATIVE
+	if (mal_has_feature(mal, MAL_FTR_CLEAR_ICINTSTAT))
+		mtdcri(SDR0, DCRN_SDR_ICINTSTAT,
+				(mfdcri(SDR0, DCRN_SDR_ICINTSTAT) | ICINTSTAT_ICTX));
+#endif
+
+	return IRQ_HANDLED;
+}
+
+static irqreturn_t mal_rxeob(int irq, void *dev_instance)
+{
+	struct mal_instance *mal = dev_instance;
+
+	u32 r = get_mal_dcrn(mal, MAL_RXEOBISR);
+
+	MAL_DBG2(mal, "rxeob %08x" NL, r);
+
+	mal_schedule_poll(mal);
+	set_mal_dcrn(mal, MAL_RXEOBISR, r);
+
+#ifdef CONFIG_PPC_DCR_NATIVE
+	if (mal_has_feature(mal, MAL_FTR_CLEAR_ICINTSTAT))
+		mtdcri(SDR0, DCRN_SDR_ICINTSTAT,
+				(mfdcri(SDR0, DCRN_SDR_ICINTSTAT) | ICINTSTAT_ICRX));
+#endif
+
+	return IRQ_HANDLED;
+}
+
+static irqreturn_t mal_txde(int irq, void *dev_instance)
+{
+	struct mal_instance *mal = dev_instance;
+
+	u32 deir = get_mal_dcrn(mal, MAL_TXDEIR);
+	set_mal_dcrn(mal, MAL_TXDEIR, deir);
+
+	MAL_DBG(mal, "txde %08x" NL, deir);
+
+	if (net_ratelimit())
+		printk(KERN_ERR
+		       "mal%d: TX descriptor error (TXDEIR = 0x%08x)\n",
+		       mal->index, deir);
+
+	return IRQ_HANDLED;
+}
+
+static irqreturn_t mal_rxde(int irq, void *dev_instance)
+{
+	struct mal_instance *mal = dev_instance;
+	struct list_head *l;
+
+	u32 deir = get_mal_dcrn(mal, MAL_RXDEIR);
+
+	MAL_DBG(mal, "rxde %08x" NL, deir);
+
+	list_for_each(l, &mal->list) {
+		struct mal_commac *mc = list_entry(l, struct mal_commac, list);
+		if (deir & mc->rx_chan_mask) {
+			set_bit(MAL_COMMAC_RX_STOPPED, &mc->flags);
+			mc->ops->rxde(mc->dev);
+		}
+	}
+
+	mal_schedule_poll(mal);
+	set_mal_dcrn(mal, MAL_RXDEIR, deir);
+
+	return IRQ_HANDLED;
+}
+
+static irqreturn_t mal_int(int irq, void *dev_instance)
+{
+	struct mal_instance *mal = dev_instance;
+	u32 esr = get_mal_dcrn(mal, MAL_ESR);
+
+	if (esr & MAL_ESR_EVB) {
+		/* descriptor error */
+		if (esr & MAL_ESR_DE) {
+			if (esr & MAL_ESR_CIDT)
+				return mal_rxde(irq, dev_instance);
+			else
+				return mal_txde(irq, dev_instance);
+		} else { /* SERR */
+			return mal_serr(irq, dev_instance);
+		}
+	}
+	return IRQ_HANDLED;
+}
+
+void mal_poll_disable(struct mal_instance *mal, struct mal_commac *commac)
+{
+	/* Spinlock-type semantics: only one caller disable poll at a time */
+	while (test_and_set_bit(MAL_COMMAC_POLL_DISABLED, &commac->flags))
+		msleep(1);
+
+	/* Synchronize with the MAL NAPI poller */
+	napi_synchronize(&mal->napi);
+}
+
+void mal_poll_enable(struct mal_instance *mal, struct mal_commac *commac)
+{
+	smp_wmb();
+	clear_bit(MAL_COMMAC_POLL_DISABLED, &commac->flags);
+
+	/* Feels better to trigger a poll here to catch up with events that
+	 * may have happened on this channel while disabled. It will most
+	 * probably be delayed until the next interrupt but that's mostly a
+	 * non-issue in the context where this is called.
+	 */
+	napi_schedule(&mal->napi);
+}
+
+static int mal_poll(struct napi_struct *napi, int budget)
+{
+	struct mal_instance *mal = container_of(napi, struct mal_instance, napi);
+	struct list_head *l;
+	int received = 0;
+	unsigned long flags;
+
+	MAL_DBG2(mal, "poll(%d)" NL, budget);
+ again:
+	/* Process TX skbs */
+	list_for_each(l, &mal->poll_list) {
+		struct mal_commac *mc =
+			list_entry(l, struct mal_commac, poll_list);
+		mc->ops->poll_tx(mc->dev);
+	}
+
+	/* Process RX skbs.
+	 *
+	 * We _might_ need something more smart here to enforce polling
+	 * fairness.
+	 */
+	list_for_each(l, &mal->poll_list) {
+		struct mal_commac *mc =
+			list_entry(l, struct mal_commac, poll_list);
+		int n;
+		if (unlikely(test_bit(MAL_COMMAC_POLL_DISABLED, &mc->flags)))
+			continue;
+		n = mc->ops->poll_rx(mc->dev, budget);
+		if (n) {
+			received += n;
+			budget -= n;
+			if (budget <= 0)
+				goto more_work; // XXX What if this is the last one ?
+		}
+	}
+
+	/* We need to disable IRQs to protect from RXDE IRQ here */
+	spin_lock_irqsave(&mal->lock, flags);
+	__napi_complete(napi);
+	mal_enable_eob_irq(mal);
+	spin_unlock_irqrestore(&mal->lock, flags);
+
+	/* Check for "rotting" packet(s) */
+	list_for_each(l, &mal->poll_list) {
+		struct mal_commac *mc =
+			list_entry(l, struct mal_commac, poll_list);
+		if (unlikely(test_bit(MAL_COMMAC_POLL_DISABLED, &mc->flags)))
+			continue;
+		if (unlikely(mc->ops->peek_rx(mc->dev) ||
+			     test_bit(MAL_COMMAC_RX_STOPPED, &mc->flags))) {
+			MAL_DBG2(mal, "rotting packet" NL);
+			if (!napi_reschedule(napi))
+				goto more_work;
+
+			spin_lock_irqsave(&mal->lock, flags);
+			mal_disable_eob_irq(mal);
+			spin_unlock_irqrestore(&mal->lock, flags);
+			goto again;
+		}
+		mc->ops->poll_tx(mc->dev);
+	}
+
+ more_work:
+	MAL_DBG2(mal, "poll() %d <- %d" NL, budget, received);
+	return received;
+}
+
+static void mal_reset(struct mal_instance *mal)
+{
+	int n = 10;
+
+	MAL_DBG(mal, "reset" NL);
+
+	set_mal_dcrn(mal, MAL_CFG, MAL_CFG_SR);
+
+	/* Wait for reset to complete (1 system clock) */
+	while ((get_mal_dcrn(mal, MAL_CFG) & MAL_CFG_SR) && n)
+		--n;
+
+	if (unlikely(!n))
+		printk(KERN_ERR "mal%d: reset timeout\n", mal->index);
+}
+
+int mal_get_regs_len(struct mal_instance *mal)
+{
+	return sizeof(struct emac_ethtool_regs_subhdr) +
+	    sizeof(struct mal_regs);
+}
+
+void *mal_dump_regs(struct mal_instance *mal, void *buf)
+{
+	struct emac_ethtool_regs_subhdr *hdr = buf;
+	struct mal_regs *regs = (struct mal_regs *)(hdr + 1);
+	int i;
+
+	hdr->version = mal->version;
+	hdr->index = mal->index;
+
+	regs->tx_count = mal->num_tx_chans;
+	regs->rx_count = mal->num_rx_chans;
+
+	regs->cfg = get_mal_dcrn(mal, MAL_CFG);
+	regs->esr = get_mal_dcrn(mal, MAL_ESR);
+	regs->ier = get_mal_dcrn(mal, MAL_IER);
+	regs->tx_casr = get_mal_dcrn(mal, MAL_TXCASR);
+	regs->tx_carr = get_mal_dcrn(mal, MAL_TXCARR);
+	regs->tx_eobisr = get_mal_dcrn(mal, MAL_TXEOBISR);
+	regs->tx_deir = get_mal_dcrn(mal, MAL_TXDEIR);
+	regs->rx_casr = get_mal_dcrn(mal, MAL_RXCASR);
+	regs->rx_carr = get_mal_dcrn(mal, MAL_RXCARR);
+	regs->rx_eobisr = get_mal_dcrn(mal, MAL_RXEOBISR);
+	regs->rx_deir = get_mal_dcrn(mal, MAL_RXDEIR);
+
+	for (i = 0; i < regs->tx_count; ++i)
+		regs->tx_ctpr[i] = get_mal_dcrn(mal, MAL_TXCTPR(i));
+
+	for (i = 0; i < regs->rx_count; ++i) {
+		regs->rx_ctpr[i] = get_mal_dcrn(mal, MAL_RXCTPR(i));
+		regs->rcbs[i] = get_mal_dcrn(mal, MAL_RCBS(i));
+	}
+	return regs + 1;
+}
+
+static int mal_probe(struct platform_device *ofdev)
+{
+	struct mal_instance *mal;
+	int err = 0, i, bd_size;
+	int index = mal_count++;
+	unsigned int dcr_base;
+	const u32 *prop;
+	u32 cfg;
+	unsigned long irqflags;
+	irq_handler_t hdlr_serr, hdlr_txde, hdlr_rxde;
+
+	mal = kzalloc(sizeof(struct mal_instance), GFP_KERNEL);
+	if (!mal)
+		return -ENOMEM;
+
+	mal->index = index;
+	mal->ofdev = ofdev;
+	mal->version = of_device_is_compatible(ofdev->dev.of_node, "ibm,mcmal2") ? 2 : 1;
+
+	MAL_DBG(mal, "probe" NL);
+
+	prop = of_get_property(ofdev->dev.of_node, "num-tx-chans", NULL);
+	if (prop == NULL) {
+		printk(KERN_ERR
+		       "mal%d: can't find MAL num-tx-chans property!\n",
+		       index);
+		err = -ENODEV;
+		goto fail;
+	}
+	mal->num_tx_chans = prop[0];
+
+	prop = of_get_property(ofdev->dev.of_node, "num-rx-chans", NULL);
+	if (prop == NULL) {
+		printk(KERN_ERR
+		       "mal%d: can't find MAL num-rx-chans property!\n",
+		       index);
+		err = -ENODEV;
+		goto fail;
+	}
+	mal->num_rx_chans = prop[0];
+
+	dcr_base = dcr_resource_start(ofdev->dev.of_node, 0);
+	if (dcr_base == 0) {
+		printk(KERN_ERR
+		       "mal%d: can't find DCR resource!\n", index);
+		err = -ENODEV;
+		goto fail;
+	}
+	mal->dcr_host = dcr_map(ofdev->dev.of_node, dcr_base, 0x100);
+	if (!DCR_MAP_OK(mal->dcr_host)) {
+		printk(KERN_ERR
+		       "mal%d: failed to map DCRs !\n", index);
+		err = -ENODEV;
+		goto fail;
+	}
+
+	if (of_device_is_compatible(ofdev->dev.of_node, "ibm,mcmal-405ez")) {
+#if defined(CONFIG_IBM_EMAC_MAL_CLR_ICINTSTAT) && \
+		defined(CONFIG_IBM_EMAC_MAL_COMMON_ERR)
+		mal->features |= (MAL_FTR_CLEAR_ICINTSTAT |
+				MAL_FTR_COMMON_ERR_INT);
+#else
+		printk(KERN_ERR "%s: Support for 405EZ not enabled!\n",
+				ofdev->dev.of_node->full_name);
+		err = -ENODEV;
+		goto fail;
+#endif
+	}
+
+	mal->txeob_irq = irq_of_parse_and_map(ofdev->dev.of_node, 0);
+	mal->rxeob_irq = irq_of_parse_and_map(ofdev->dev.of_node, 1);
+	mal->serr_irq = irq_of_parse_and_map(ofdev->dev.of_node, 2);
+
+	if (mal_has_feature(mal, MAL_FTR_COMMON_ERR_INT)) {
+		mal->txde_irq = mal->rxde_irq = mal->serr_irq;
+	} else {
+		mal->txde_irq = irq_of_parse_and_map(ofdev->dev.of_node, 3);
+		mal->rxde_irq = irq_of_parse_and_map(ofdev->dev.of_node, 4);
+	}
+
+	if (mal->txeob_irq == NO_IRQ || mal->rxeob_irq == NO_IRQ ||
+	    mal->serr_irq == NO_IRQ || mal->txde_irq == NO_IRQ ||
+	    mal->rxde_irq == NO_IRQ) {
+		printk(KERN_ERR
+		       "mal%d: failed to map interrupts !\n", index);
+		err = -ENODEV;
+		goto fail_unmap;
+	}
+
+	INIT_LIST_HEAD(&mal->poll_list);
+	INIT_LIST_HEAD(&mal->list);
+	spin_lock_init(&mal->lock);
+
+	init_dummy_netdev(&mal->dummy_dev);
+
+	netif_napi_add(&mal->dummy_dev, &mal->napi, mal_poll,
+		       CONFIG_IBM_EMAC_POLL_WEIGHT);
+
+	/* Load power-on reset defaults */
+	mal_reset(mal);
+
+	/* Set the MAL configuration register */
+	cfg = (mal->version == 2) ? MAL2_CFG_DEFAULT : MAL1_CFG_DEFAULT;
+	cfg |= MAL_CFG_PLBB | MAL_CFG_OPBBL | MAL_CFG_LEA;
+
+	/* Current Axon is not happy with priority being non-0, it can
+	 * deadlock, fix it up here
+	 */
+	if (of_device_is_compatible(ofdev->dev.of_node, "ibm,mcmal-axon"))
+		cfg &= ~(MAL2_CFG_RPP_10 | MAL2_CFG_WPP_10);
+
+	/* Apply configuration */
+	set_mal_dcrn(mal, MAL_CFG, cfg);
+
+	/* Allocate space for BD rings */
+	BUG_ON(mal->num_tx_chans <= 0 || mal->num_tx_chans > 32);
+	BUG_ON(mal->num_rx_chans <= 0 || mal->num_rx_chans > 32);
+
+	bd_size = sizeof(struct mal_descriptor) *
+		(NUM_TX_BUFF * mal->num_tx_chans +
+		 NUM_RX_BUFF * mal->num_rx_chans);
+	mal->bd_virt = dma_zalloc_coherent(&ofdev->dev, bd_size, &mal->bd_dma,
+					   GFP_KERNEL);
+	if (mal->bd_virt == NULL) {
+		err = -ENOMEM;
+		goto fail_unmap;
+	}
+
+	for (i = 0; i < mal->num_tx_chans; ++i)
+		set_mal_dcrn(mal, MAL_TXCTPR(i), mal->bd_dma +
+			     sizeof(struct mal_descriptor) *
+			     mal_tx_bd_offset(mal, i));
+
+	for (i = 0; i < mal->num_rx_chans; ++i)
+		set_mal_dcrn(mal, MAL_RXCTPR(i), mal->bd_dma +
+			     sizeof(struct mal_descriptor) *
+			     mal_rx_bd_offset(mal, i));
+
+	if (mal_has_feature(mal, MAL_FTR_COMMON_ERR_INT)) {
+		irqflags = IRQF_SHARED;
+		hdlr_serr = hdlr_txde = hdlr_rxde = mal_int;
+	} else {
+		irqflags = 0;
+		hdlr_serr = mal_serr;
+		hdlr_txde = mal_txde;
+		hdlr_rxde = mal_rxde;
+	}
+
+	err = request_irq(mal->serr_irq, hdlr_serr, irqflags, "MAL SERR", mal);
+	if (err)
+		goto fail2;
+	err = request_irq(mal->txde_irq, hdlr_txde, irqflags, "MAL TX DE", mal);
+	if (err)
+		goto fail3;
+	err = request_irq(mal->txeob_irq, mal_txeob, 0, "MAL TX EOB", mal);
+	if (err)
+		goto fail4;
+	err = request_irq(mal->rxde_irq, hdlr_rxde, irqflags, "MAL RX DE", mal);
+	if (err)
+		goto fail5;
+	err = request_irq(mal->rxeob_irq, mal_rxeob, 0, "MAL RX EOB", mal);
+	if (err)
+		goto fail6;
+
+	/* Enable all MAL SERR interrupt sources */
+	set_mal_dcrn(mal, MAL_IER, MAL_IER_EVENTS);
+
+	/* Enable EOB interrupt */
+	mal_enable_eob_irq(mal);
+
+	printk(KERN_INFO
+	       "MAL v%d %s, %d TX channels, %d RX channels\n",
+	       mal->version, ofdev->dev.of_node->full_name,
+	       mal->num_tx_chans, mal->num_rx_chans);
+
+	/* Advertise this instance to the rest of the world */
+	wmb();
+	platform_set_drvdata(ofdev, mal);
+
+	mal_dbg_register(mal);
+
+	return 0;
+
+ fail6:
+	free_irq(mal->rxde_irq, mal);
+ fail5:
+	free_irq(mal->txeob_irq, mal);
+ fail4:
+	free_irq(mal->txde_irq, mal);
+ fail3:
+	free_irq(mal->serr_irq, mal);
+ fail2:
+	dma_free_coherent(&ofdev->dev, bd_size, mal->bd_virt, mal->bd_dma);
+ fail_unmap:
+	dcr_unmap(mal->dcr_host, 0x100);
+ fail:
+	kfree(mal);
+
+	return err;
+}
+
+static int mal_remove(struct platform_device *ofdev)
+{
+	struct mal_instance *mal = platform_get_drvdata(ofdev);
+
+	MAL_DBG(mal, "remove" NL);
+
+	/* Synchronize with scheduled polling */
+	napi_disable(&mal->napi);
+
+	if (!list_empty(&mal->list))
+		/* This is *very* bad */
+		WARN(1, KERN_EMERG
+		       "mal%d: commac list is not empty on remove!\n",
+		       mal->index);
+
+	free_irq(mal->serr_irq, mal);
+	free_irq(mal->txde_irq, mal);
+	free_irq(mal->txeob_irq, mal);
+	free_irq(mal->rxde_irq, mal);
+	free_irq(mal->rxeob_irq, mal);
+
+	mal_reset(mal);
+
+	mal_dbg_unregister(mal);
+
+	dma_free_coherent(&ofdev->dev,
+			  sizeof(struct mal_descriptor) *
+			  (NUM_TX_BUFF * mal->num_tx_chans +
+			   NUM_RX_BUFF * mal->num_rx_chans), mal->bd_virt,
+			  mal->bd_dma);
+	kfree(mal);
+
+	return 0;
+}
+
+static const struct of_device_id mal_platform_match[] =
+{
+	{
+		.compatible	= "ibm,mcmal",
+	},
+	{
+		.compatible	= "ibm,mcmal2",
+	},
+	/* Backward compat */
+	{
+		.type		= "mcmal-dma",
+		.compatible	= "ibm,mcmal",
+	},
+	{
+		.type		= "mcmal-dma",
+		.compatible	= "ibm,mcmal2",
+	},
+	{},
+};
+
+static struct platform_driver mal_of_driver = {
+	.driver = {
+		.name = "mcmal",
+		.of_match_table = mal_platform_match,
+	},
+	.probe = mal_probe,
+	.remove = mal_remove,
+};
+
+int __init mal_init(void)
+{
+	return platform_driver_register(&mal_of_driver);
+}
+
+void mal_exit(void)
+{
+	platform_driver_unregister(&mal_of_driver);
+}
diff --git a/drivers/net/ethernet/ibm/emac/mal.h b/drivers/net/ethernet/ibm/emac/mal.h
new file mode 100644
index 000000000..eeade2ea8
--- /dev/null
+++ b/drivers/net/ethernet/ibm/emac/mal.h
@@ -0,0 +1,312 @@
+/*
+ * drivers/net/ethernet/ibm/emac/mal.h
+ *
+ * Memory Access Layer (MAL) support
+ *
+ * Copyright 2007 Benjamin Herrenschmidt, IBM Corp.
+ *                <benh@kernel.crashing.org>
+ *
+ * Based on the arch/ppc version of the driver:
+ *
+ * Copyright (c) 2004, 2005 Zultys Technologies.
+ * Eugene Surovegin <eugene.surovegin@zultys.com> or <ebs@ebshome.net>
+ *
+ * Based on original work by
+ *      Armin Kuster <akuster@mvista.com>
+ *      Copyright 2002 MontaVista Softare Inc.
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ *
+ */
+#ifndef __IBM_NEWEMAC_MAL_H
+#define __IBM_NEWEMAC_MAL_H
+
+/*
+ * There are some variations on the MAL, we express them in this driver as
+ * MAL Version 1 and 2 though that doesn't match any IBM terminology.
+ *
+ * We call MAL 1 the version in 405GP, 405GPR, 405EP, 440EP, 440GR and
+ * NP405H.
+ *
+ * We call MAL 2 the version in 440GP, 440GX, 440SP, 440SPE and Axon
+ *
+ * The driver expects a "version" property in the emac node containing
+ * a number 1 or 2. New device-trees for EMAC capable platforms are thus
+ * required to include that when porting to arch/powerpc.
+ */
+
+/* MALx DCR registers */
+#define	MAL_CFG			0x00
+#define	  MAL_CFG_SR		0x80000000
+#define   MAL_CFG_PLBB		0x00004000
+#define   MAL_CFG_OPBBL		0x00000080
+#define   MAL_CFG_EOPIE		0x00000004
+#define   MAL_CFG_LEA		0x00000002
+#define   MAL_CFG_SD		0x00000001
+
+/* MAL V1 CFG bits */
+#define   MAL1_CFG_PLBP_MASK	0x00c00000
+#define   MAL1_CFG_PLBP_10	0x00800000
+#define   MAL1_CFG_GA		0x00200000
+#define   MAL1_CFG_OA		0x00100000
+#define   MAL1_CFG_PLBLE	0x00080000
+#define   MAL1_CFG_PLBT_MASK	0x00078000
+#define   MAL1_CFG_DEFAULT	(MAL1_CFG_PLBP_10 | MAL1_CFG_PLBT_MASK)
+
+/* MAL V2 CFG bits */
+#define   MAL2_CFG_RPP_MASK	0x00c00000
+#define   MAL2_CFG_RPP_10	0x00800000
+#define   MAL2_CFG_RMBS_MASK	0x00300000
+#define   MAL2_CFG_WPP_MASK	0x000c0000
+#define   MAL2_CFG_WPP_10	0x00080000
+#define   MAL2_CFG_WMBS_MASK	0x00030000
+#define   MAL2_CFG_PLBLE	0x00008000
+#define   MAL2_CFG_DEFAULT	(MAL2_CFG_RMBS_MASK | MAL2_CFG_WMBS_MASK | \
+				 MAL2_CFG_RPP_10 | MAL2_CFG_WPP_10)
+
+#define MAL_ESR			0x01
+#define   MAL_ESR_EVB		0x80000000
+#define   MAL_ESR_CIDT		0x40000000
+#define   MAL_ESR_CID_MASK	0x3e000000
+#define   MAL_ESR_CID_SHIFT	25
+#define   MAL_ESR_DE		0x00100000
+#define   MAL_ESR_OTE		0x00040000
+#define   MAL_ESR_OSE		0x00020000
+#define   MAL_ESR_PEIN		0x00010000
+#define   MAL_ESR_DEI		0x00000010
+#define   MAL_ESR_OTEI		0x00000004
+#define   MAL_ESR_OSEI		0x00000002
+#define   MAL_ESR_PBEI		0x00000001
+
+/* MAL V1 ESR bits */
+#define   MAL1_ESR_ONE		0x00080000
+#define   MAL1_ESR_ONEI		0x00000008
+
+/* MAL V2 ESR bits */
+#define   MAL2_ESR_PTE		0x00800000
+#define   MAL2_ESR_PRE		0x00400000
+#define   MAL2_ESR_PWE		0x00200000
+#define   MAL2_ESR_PTEI		0x00000080
+#define   MAL2_ESR_PREI		0x00000040
+#define   MAL2_ESR_PWEI		0x00000020
+
+
+#define MAL_IER			0x02
+/* MAL IER bits */
+#define   MAL_IER_DE		0x00000010
+#define   MAL_IER_OTE		0x00000004
+#define   MAL_IER_OE		0x00000002
+#define   MAL_IER_PE		0x00000001
+
+/* PLB read/write/timeout errors */
+#define   MAL_IER_PTE		0x00000080
+#define   MAL_IER_PRE		0x00000040
+#define   MAL_IER_PWE		0x00000020
+
+#define   MAL_IER_SOC_EVENTS	(MAL_IER_PTE | MAL_IER_PRE | MAL_IER_PWE)
+#define   MAL_IER_EVENTS	(MAL_IER_SOC_EVENTS | MAL_IER_DE | \
+				 MAL_IER_OTE | MAL_IER_OE | MAL_IER_PE)
+
+#define MAL_TXCASR		0x04
+#define MAL_TXCARR		0x05
+#define MAL_TXEOBISR		0x06
+#define MAL_TXDEIR		0x07
+#define MAL_RXCASR		0x10
+#define MAL_RXCARR		0x11
+#define MAL_RXEOBISR		0x12
+#define MAL_RXDEIR		0x13
+#define MAL_TXCTPR(n)		((n) + 0x20)
+#define MAL_RXCTPR(n)		((n) + 0x40)
+#define MAL_RCBS(n)		((n) + 0x60)
+
+/* In reality MAL can handle TX buffers up to 4095 bytes long,
+ * but this isn't a good round number :) 		 --ebs
+ */
+#define MAL_MAX_TX_SIZE		4080
+#define MAL_MAX_RX_SIZE		4080
+
+static inline int mal_rx_size(int len)
+{
+	len = (len + 0xf) & ~0xf;
+	return len > MAL_MAX_RX_SIZE ? MAL_MAX_RX_SIZE : len;
+}
+
+static inline int mal_tx_chunks(int len)
+{
+	return (len + MAL_MAX_TX_SIZE - 1) / MAL_MAX_TX_SIZE;
+}
+
+#define MAL_CHAN_MASK(n)	(0x80000000 >> (n))
+
+/* MAL Buffer Descriptor structure */
+struct mal_descriptor {
+	u16 ctrl;		/* MAL / Commac status control bits */
+	u16 data_len;		/* Max length is 4K-1 (12 bits)     */
+	u32 data_ptr;		/* pointer to actual data buffer    */
+};
+
+/* the following defines are for the MadMAL status and control registers. */
+/* MADMAL transmit and receive status/control bits  */
+#define MAL_RX_CTRL_EMPTY	0x8000
+#define MAL_RX_CTRL_WRAP	0x4000
+#define MAL_RX_CTRL_CM		0x2000
+#define MAL_RX_CTRL_LAST	0x1000
+#define MAL_RX_CTRL_FIRST	0x0800
+#define MAL_RX_CTRL_INTR	0x0400
+#define MAL_RX_CTRL_SINGLE	(MAL_RX_CTRL_LAST | MAL_RX_CTRL_FIRST)
+#define MAL_IS_SINGLE_RX(ctrl)	(((ctrl) & MAL_RX_CTRL_SINGLE) == MAL_RX_CTRL_SINGLE)
+
+#define MAL_TX_CTRL_READY	0x8000
+#define MAL_TX_CTRL_WRAP	0x4000
+#define MAL_TX_CTRL_CM		0x2000
+#define MAL_TX_CTRL_LAST	0x1000
+#define MAL_TX_CTRL_INTR	0x0400
+
+struct mal_commac_ops {
+	void	(*poll_tx) (void *dev);
+	int	(*poll_rx) (void *dev, int budget);
+	int	(*peek_rx) (void *dev);
+	void	(*rxde) (void *dev);
+};
+
+struct mal_commac {
+	struct mal_commac_ops	*ops;
+	void			*dev;
+	struct list_head	poll_list;
+	long       		flags;
+#define MAL_COMMAC_RX_STOPPED		0
+#define MAL_COMMAC_POLL_DISABLED	1
+	u32			tx_chan_mask;
+	u32			rx_chan_mask;
+	struct list_head	list;
+};
+
+struct mal_instance {
+	int			version;
+	dcr_host_t		dcr_host;
+
+	int			num_tx_chans;	/* Number of TX channels */
+	int			num_rx_chans;	/* Number of RX channels */
+	int 			txeob_irq;	/* TX End Of Buffer IRQ  */
+	int 			rxeob_irq;	/* RX End Of Buffer IRQ  */
+	int			txde_irq;	/* TX Descriptor Error IRQ */
+	int			rxde_irq;	/* RX Descriptor Error IRQ */
+	int			serr_irq;	/* MAL System Error IRQ    */
+
+	struct list_head	poll_list;
+	struct napi_struct	napi;
+
+	struct list_head	list;
+	u32			tx_chan_mask;
+	u32			rx_chan_mask;
+
+	dma_addr_t		bd_dma;
+	struct mal_descriptor	*bd_virt;
+
+	struct platform_device	*ofdev;
+	int			index;
+	spinlock_t		lock;
+
+	struct net_device	dummy_dev;
+
+	unsigned int features;
+};
+
+static inline u32 get_mal_dcrn(struct mal_instance *mal, int reg)
+{
+	return dcr_read(mal->dcr_host, reg);
+}
+
+static inline void set_mal_dcrn(struct mal_instance *mal, int reg, u32 val)
+{
+	dcr_write(mal->dcr_host, reg, val);
+}
+
+/* Features of various MAL implementations */
+
+/* Set if you have interrupt coalescing and you have to clear the SDR
+ * register for TXEOB and RXEOB interrupts to work
+ */
+#define MAL_FTR_CLEAR_ICINTSTAT	0x00000001
+
+/* Set if your MAL has SERR, TXDE, and RXDE OR'd into a single UIC
+ * interrupt
+ */
+#define MAL_FTR_COMMON_ERR_INT	0x00000002
+
+enum {
+	MAL_FTRS_ALWAYS = 0,
+
+	MAL_FTRS_POSSIBLE =
+#ifdef CONFIG_IBM_EMAC_MAL_CLR_ICINTSTAT
+		MAL_FTR_CLEAR_ICINTSTAT |
+#endif
+#ifdef CONFIG_IBM_EMAC_MAL_COMMON_ERR
+		MAL_FTR_COMMON_ERR_INT |
+#endif
+		0,
+};
+
+static inline int mal_has_feature(struct mal_instance *dev,
+		unsigned long feature)
+{
+	return (MAL_FTRS_ALWAYS & feature) ||
+		(MAL_FTRS_POSSIBLE & dev->features & feature);
+}
+
+/* Register MAL devices */
+int mal_init(void);
+void mal_exit(void);
+
+int mal_register_commac(struct mal_instance *mal,
+			struct mal_commac *commac);
+void mal_unregister_commac(struct mal_instance *mal,
+			   struct mal_commac *commac);
+int mal_set_rcbs(struct mal_instance *mal, int channel, unsigned long size);
+
+/* Returns BD ring offset for a particular channel
+   (in 'struct mal_descriptor' elements)
+*/
+int mal_tx_bd_offset(struct mal_instance *mal, int channel);
+int mal_rx_bd_offset(struct mal_instance *mal, int channel);
+
+void mal_enable_tx_channel(struct mal_instance *mal, int channel);
+void mal_disable_tx_channel(struct mal_instance *mal, int channel);
+void mal_enable_rx_channel(struct mal_instance *mal, int channel);
+void mal_disable_rx_channel(struct mal_instance *mal, int channel);
+
+void mal_poll_disable(struct mal_instance *mal, struct mal_commac *commac);
+void mal_poll_enable(struct mal_instance *mal, struct mal_commac *commac);
+
+/* Add/remove EMAC to/from MAL polling list */
+void mal_poll_add(struct mal_instance *mal, struct mal_commac *commac);
+void mal_poll_del(struct mal_instance *mal, struct mal_commac *commac);
+
+/* Ethtool MAL registers */
+struct mal_regs {
+	u32 tx_count;
+	u32 rx_count;
+
+	u32 cfg;
+	u32 esr;
+	u32 ier;
+	u32 tx_casr;
+	u32 tx_carr;
+	u32 tx_eobisr;
+	u32 tx_deir;
+	u32 rx_casr;
+	u32 rx_carr;
+	u32 rx_eobisr;
+	u32 rx_deir;
+	u32 tx_ctpr[32];
+	u32 rx_ctpr[32];
+	u32 rcbs[32];
+};
+
+int mal_get_regs_len(struct mal_instance *mal);
+void *mal_dump_regs(struct mal_instance *mal, void *buf);
+
+#endif /* __IBM_NEWEMAC_MAL_H */
diff --git a/drivers/net/ethernet/ibm/emac/phy.c b/drivers/net/ethernet/ibm/emac/phy.c
new file mode 100644
index 000000000..d3b9d1033
--- /dev/null
+++ b/drivers/net/ethernet/ibm/emac/phy.c
@@ -0,0 +1,541 @@
+/*
+ * drivers/net/ethernet/ibm/emac/phy.c
+ *
+ * Driver for PowerPC 4xx on-chip ethernet controller, PHY support.
+ * Borrowed from sungem_phy.c, though I only kept the generic MII
+ * driver for now.
+ *
+ * This file should be shared with other drivers or eventually
+ * merged as the "low level" part of miilib
+ *
+ * Copyright 2007 Benjamin Herrenschmidt, IBM Corp.
+ *                <benh@kernel.crashing.org>
+ *
+ * Based on the arch/ppc version of the driver:
+ *
+ * (c) 2003, Benjamin Herrenscmidt (benh@kernel.crashing.org)
+ * (c) 2004-2005, Eugene Surovegin <ebs@ebshome.net>
+ *
+ */
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/types.h>
+#include <linux/netdevice.h>
+#include <linux/mii.h>
+#include <linux/ethtool.h>
+#include <linux/delay.h>
+
+#include "emac.h"
+#include "phy.h"
+
+#define phy_read _phy_read
+#define phy_write _phy_write
+
+static inline int _phy_read(struct mii_phy *phy, int reg)
+{
+	return phy->mdio_read(phy->dev, phy->address, reg);
+}
+
+static inline void _phy_write(struct mii_phy *phy, int reg, int val)
+{
+	phy->mdio_write(phy->dev, phy->address, reg, val);
+}
+
+static inline int gpcs_phy_read(struct mii_phy *phy, int reg)
+{
+	return phy->mdio_read(phy->dev, phy->gpcs_address, reg);
+}
+
+static inline void gpcs_phy_write(struct mii_phy *phy, int reg, int val)
+{
+	phy->mdio_write(phy->dev, phy->gpcs_address, reg, val);
+}
+
+int emac_mii_reset_phy(struct mii_phy *phy)
+{
+	int val;
+	int limit = 10000;
+
+	val = phy_read(phy, MII_BMCR);
+	val &= ~(BMCR_ISOLATE | BMCR_ANENABLE);
+	val |= BMCR_RESET;
+	phy_write(phy, MII_BMCR, val);
+
+	udelay(300);
+
+	while (--limit) {
+		val = phy_read(phy, MII_BMCR);
+		if (val >= 0 && (val & BMCR_RESET) == 0)
+			break;
+		udelay(10);
+	}
+	if ((val & BMCR_ISOLATE) && limit > 0)
+		phy_write(phy, MII_BMCR, val & ~BMCR_ISOLATE);
+
+	return limit <= 0;
+}
+
+int emac_mii_reset_gpcs(struct mii_phy *phy)
+{
+	int val;
+	int limit = 10000;
+
+	val = gpcs_phy_read(phy, MII_BMCR);
+	val &= ~(BMCR_ISOLATE | BMCR_ANENABLE);
+	val |= BMCR_RESET;
+	gpcs_phy_write(phy, MII_BMCR, val);
+
+	udelay(300);
+
+	while (--limit) {
+		val = gpcs_phy_read(phy, MII_BMCR);
+		if (val >= 0 && (val & BMCR_RESET) == 0)
+			break;
+		udelay(10);
+	}
+	if ((val & BMCR_ISOLATE) && limit > 0)
+		gpcs_phy_write(phy, MII_BMCR, val & ~BMCR_ISOLATE);
+
+	if (limit > 0 && phy->mode == PHY_MODE_SGMII) {
+		/* Configure GPCS interface to recommended setting for SGMII */
+		gpcs_phy_write(phy, 0x04, 0x8120); /* AsymPause, FDX */
+		gpcs_phy_write(phy, 0x07, 0x2801); /* msg_pg, toggle */
+		gpcs_phy_write(phy, 0x00, 0x0140); /* 1Gbps, FDX     */
+	}
+
+	return limit <= 0;
+}
+
+static int genmii_setup_aneg(struct mii_phy *phy, u32 advertise)
+{
+	int ctl, adv;
+
+	phy->autoneg = AUTONEG_ENABLE;
+	phy->speed = SPEED_10;
+	phy->duplex = DUPLEX_HALF;
+	phy->pause = phy->asym_pause = 0;
+	phy->advertising = advertise;
+
+	ctl = phy_read(phy, MII_BMCR);
+	if (ctl < 0)
+		return ctl;
+	ctl &= ~(BMCR_FULLDPLX | BMCR_SPEED100 | BMCR_SPEED1000 | BMCR_ANENABLE);
+
+	/* First clear the PHY */
+	phy_write(phy, MII_BMCR, ctl);
+
+	/* Setup standard advertise */
+	adv = phy_read(phy, MII_ADVERTISE);
+	if (adv < 0)
+		return adv;
+	adv &= ~(ADVERTISE_ALL | ADVERTISE_100BASE4 | ADVERTISE_PAUSE_CAP |
+		 ADVERTISE_PAUSE_ASYM);
+	if (advertise & ADVERTISED_10baseT_Half)
+		adv |= ADVERTISE_10HALF;
+	if (advertise & ADVERTISED_10baseT_Full)
+		adv |= ADVERTISE_10FULL;
+	if (advertise & ADVERTISED_100baseT_Half)
+		adv |= ADVERTISE_100HALF;
+	if (advertise & ADVERTISED_100baseT_Full)
+		adv |= ADVERTISE_100FULL;
+	if (advertise & ADVERTISED_Pause)
+		adv |= ADVERTISE_PAUSE_CAP;
+	if (advertise & ADVERTISED_Asym_Pause)
+		adv |= ADVERTISE_PAUSE_ASYM;
+	phy_write(phy, MII_ADVERTISE, adv);
+
+	if (phy->features &
+	    (SUPPORTED_1000baseT_Full | SUPPORTED_1000baseT_Half)) {
+		adv = phy_read(phy, MII_CTRL1000);
+		if (adv < 0)
+			return adv;
+		adv &= ~(ADVERTISE_1000FULL | ADVERTISE_1000HALF);
+		if (advertise & ADVERTISED_1000baseT_Full)
+			adv |= ADVERTISE_1000FULL;
+		if (advertise & ADVERTISED_1000baseT_Half)
+			adv |= ADVERTISE_1000HALF;
+		phy_write(phy, MII_CTRL1000, adv);
+	}
+
+	/* Start/Restart aneg */
+	ctl = phy_read(phy, MII_BMCR);
+	ctl |= (BMCR_ANENABLE | BMCR_ANRESTART);
+	phy_write(phy, MII_BMCR, ctl);
+
+	return 0;
+}
+
+static int genmii_setup_forced(struct mii_phy *phy, int speed, int fd)
+{
+	int ctl;
+
+	phy->autoneg = AUTONEG_DISABLE;
+	phy->speed = speed;
+	phy->duplex = fd;
+	phy->pause = phy->asym_pause = 0;
+
+	ctl = phy_read(phy, MII_BMCR);
+	if (ctl < 0)
+		return ctl;
+	ctl &= ~(BMCR_FULLDPLX | BMCR_SPEED100 | BMCR_SPEED1000 | BMCR_ANENABLE);
+
+	/* First clear the PHY */
+	phy_write(phy, MII_BMCR, ctl | BMCR_RESET);
+
+	/* Select speed & duplex */
+	switch (speed) {
+	case SPEED_10:
+		break;
+	case SPEED_100:
+		ctl |= BMCR_SPEED100;
+		break;
+	case SPEED_1000:
+		ctl |= BMCR_SPEED1000;
+		break;
+	default:
+		return -EINVAL;
+	}
+	if (fd == DUPLEX_FULL)
+		ctl |= BMCR_FULLDPLX;
+	phy_write(phy, MII_BMCR, ctl);
+
+	return 0;
+}
+
+static int genmii_poll_link(struct mii_phy *phy)
+{
+	int status;
+
+	/* Clear latched value with dummy read */
+	phy_read(phy, MII_BMSR);
+	status = phy_read(phy, MII_BMSR);
+	if (status < 0 || (status & BMSR_LSTATUS) == 0)
+		return 0;
+	if (phy->autoneg == AUTONEG_ENABLE && !(status & BMSR_ANEGCOMPLETE))
+		return 0;
+	return 1;
+}
+
+static int genmii_read_link(struct mii_phy *phy)
+{
+	if (phy->autoneg == AUTONEG_ENABLE) {
+		int glpa = 0;
+		int lpa = phy_read(phy, MII_LPA) & phy_read(phy, MII_ADVERTISE);
+		if (lpa < 0)
+			return lpa;
+
+		if (phy->features &
+		    (SUPPORTED_1000baseT_Full | SUPPORTED_1000baseT_Half)) {
+			int adv = phy_read(phy, MII_CTRL1000);
+			glpa = phy_read(phy, MII_STAT1000);
+
+			if (glpa < 0 || adv < 0)
+				return adv;
+
+			glpa &= adv << 2;
+		}
+
+		phy->speed = SPEED_10;
+		phy->duplex = DUPLEX_HALF;
+		phy->pause = phy->asym_pause = 0;
+
+		if (glpa & (LPA_1000FULL | LPA_1000HALF)) {
+			phy->speed = SPEED_1000;
+			if (glpa & LPA_1000FULL)
+				phy->duplex = DUPLEX_FULL;
+		} else if (lpa & (LPA_100FULL | LPA_100HALF)) {
+			phy->speed = SPEED_100;
+			if (lpa & LPA_100FULL)
+				phy->duplex = DUPLEX_FULL;
+		} else if (lpa & LPA_10FULL)
+			phy->duplex = DUPLEX_FULL;
+
+		if (phy->duplex == DUPLEX_FULL) {
+			phy->pause = lpa & LPA_PAUSE_CAP ? 1 : 0;
+			phy->asym_pause = lpa & LPA_PAUSE_ASYM ? 1 : 0;
+		}
+	} else {
+		int bmcr = phy_read(phy, MII_BMCR);
+		if (bmcr < 0)
+			return bmcr;
+
+		if (bmcr & BMCR_FULLDPLX)
+			phy->duplex = DUPLEX_FULL;
+		else
+			phy->duplex = DUPLEX_HALF;
+		if (bmcr & BMCR_SPEED1000)
+			phy->speed = SPEED_1000;
+		else if (bmcr & BMCR_SPEED100)
+			phy->speed = SPEED_100;
+		else
+			phy->speed = SPEED_10;
+
+		phy->pause = phy->asym_pause = 0;
+	}
+	return 0;
+}
+
+/* Generic implementation for most 10/100/1000 PHYs */
+static struct mii_phy_ops generic_phy_ops = {
+	.setup_aneg	= genmii_setup_aneg,
+	.setup_forced	= genmii_setup_forced,
+	.poll_link	= genmii_poll_link,
+	.read_link	= genmii_read_link
+};
+
+static struct mii_phy_def genmii_phy_def = {
+	.phy_id		= 0x00000000,
+	.phy_id_mask	= 0x00000000,
+	.name		= "Generic MII",
+	.ops		= &generic_phy_ops
+};
+
+/* CIS8201 */
+#define MII_CIS8201_10BTCSR	0x16
+#define  TENBTCSR_ECHO_DISABLE	0x2000
+#define MII_CIS8201_EPCR	0x17
+#define  EPCR_MODE_MASK		0x3000
+#define  EPCR_GMII_MODE		0x0000
+#define  EPCR_RGMII_MODE	0x1000
+#define  EPCR_TBI_MODE		0x2000
+#define  EPCR_RTBI_MODE		0x3000
+#define MII_CIS8201_ACSR	0x1c
+#define  ACSR_PIN_PRIO_SELECT	0x0004
+
+static int cis8201_init(struct mii_phy *phy)
+{
+	int epcr;
+
+	epcr = phy_read(phy, MII_CIS8201_EPCR);
+	if (epcr < 0)
+		return epcr;
+
+	epcr &= ~EPCR_MODE_MASK;
+
+	switch (phy->mode) {
+	case PHY_MODE_TBI:
+		epcr |= EPCR_TBI_MODE;
+		break;
+	case PHY_MODE_RTBI:
+		epcr |= EPCR_RTBI_MODE;
+		break;
+	case PHY_MODE_GMII:
+		epcr |= EPCR_GMII_MODE;
+		break;
+	case PHY_MODE_RGMII:
+	default:
+		epcr |= EPCR_RGMII_MODE;
+	}
+
+	phy_write(phy, MII_CIS8201_EPCR, epcr);
+
+	/* MII regs override strap pins */
+	phy_write(phy, MII_CIS8201_ACSR,
+		  phy_read(phy, MII_CIS8201_ACSR) | ACSR_PIN_PRIO_SELECT);
+
+	/* Disable TX_EN -> CRS echo mode, otherwise 10/HDX doesn't work */
+	phy_write(phy, MII_CIS8201_10BTCSR,
+		  phy_read(phy, MII_CIS8201_10BTCSR) | TENBTCSR_ECHO_DISABLE);
+
+	return 0;
+}
+
+static struct mii_phy_ops cis8201_phy_ops = {
+	.init		= cis8201_init,
+	.setup_aneg	= genmii_setup_aneg,
+	.setup_forced	= genmii_setup_forced,
+	.poll_link	= genmii_poll_link,
+	.read_link	= genmii_read_link
+};
+
+static struct mii_phy_def cis8201_phy_def = {
+	.phy_id		= 0x000fc410,
+	.phy_id_mask	= 0x000ffff0,
+	.name		= "CIS8201 Gigabit Ethernet",
+	.ops		= &cis8201_phy_ops
+};
+
+static struct mii_phy_def bcm5248_phy_def = {
+
+	.phy_id		= 0x0143bc00,
+	.phy_id_mask	= 0x0ffffff0,
+	.name		= "BCM5248 10/100 SMII Ethernet",
+	.ops		= &generic_phy_ops
+};
+
+static int m88e1111_init(struct mii_phy *phy)
+{
+	pr_debug("%s: Marvell 88E1111 Ethernet\n", __func__);
+	phy_write(phy, 0x14, 0x0ce3);
+	phy_write(phy, 0x18, 0x4101);
+	phy_write(phy, 0x09, 0x0e00);
+	phy_write(phy, 0x04, 0x01e1);
+	phy_write(phy, 0x00, 0x9140);
+	phy_write(phy, 0x00, 0x1140);
+
+	return  0;
+}
+
+static int m88e1112_init(struct mii_phy *phy)
+{
+	/*
+	 * Marvell 88E1112 PHY needs to have the SGMII MAC
+	 * interace (page 2) properly configured to
+	 * communicate with the 460EX/GT GPCS interface.
+	 */
+
+	u16 reg_short;
+
+	pr_debug("%s: Marvell 88E1112 Ethernet\n", __func__);
+
+	/* Set access to Page 2 */
+	phy_write(phy, 0x16, 0x0002);
+
+	phy_write(phy, 0x00, 0x0040); /* 1Gbps */
+	reg_short = (u16)(phy_read(phy, 0x1a));
+	reg_short |= 0x8000; /* bypass Auto-Negotiation */
+	phy_write(phy, 0x1a, reg_short);
+	emac_mii_reset_phy(phy); /* reset MAC interface */
+
+	/* Reset access to Page 0 */
+	phy_write(phy, 0x16, 0x0000);
+
+	return  0;
+}
+
+static int et1011c_init(struct mii_phy *phy)
+{
+	u16 reg_short;
+
+	reg_short = (u16)(phy_read(phy, 0x16));
+	reg_short &= ~(0x7);
+	reg_short |= 0x6;	/* RGMII Trace Delay*/
+	phy_write(phy, 0x16, reg_short);
+
+	reg_short = (u16)(phy_read(phy, 0x17));
+	reg_short &= ~(0x40);
+	phy_write(phy, 0x17, reg_short);
+
+	phy_write(phy, 0x1c, 0x74f0);
+	return 0;
+}
+
+static struct mii_phy_ops et1011c_phy_ops = {
+	.init		= et1011c_init,
+	.setup_aneg	= genmii_setup_aneg,
+	.setup_forced	= genmii_setup_forced,
+	.poll_link	= genmii_poll_link,
+	.read_link	= genmii_read_link
+};
+
+static struct mii_phy_def et1011c_phy_def = {
+	.phy_id		= 0x0282f000,
+	.phy_id_mask	= 0x0fffff00,
+	.name		= "ET1011C Gigabit Ethernet",
+	.ops		= &et1011c_phy_ops
+};
+
+
+
+
+
+static struct mii_phy_ops m88e1111_phy_ops = {
+	.init		= m88e1111_init,
+	.setup_aneg	= genmii_setup_aneg,
+	.setup_forced	= genmii_setup_forced,
+	.poll_link	= genmii_poll_link,
+	.read_link	= genmii_read_link
+};
+
+static struct mii_phy_def m88e1111_phy_def = {
+
+	.phy_id		= 0x01410CC0,
+	.phy_id_mask	= 0x0ffffff0,
+	.name		= "Marvell 88E1111 Ethernet",
+	.ops		= &m88e1111_phy_ops,
+};
+
+static struct mii_phy_ops m88e1112_phy_ops = {
+	.init		= m88e1112_init,
+	.setup_aneg	= genmii_setup_aneg,
+	.setup_forced	= genmii_setup_forced,
+	.poll_link	= genmii_poll_link,
+	.read_link	= genmii_read_link
+};
+
+static struct mii_phy_def m88e1112_phy_def = {
+	.phy_id		= 0x01410C90,
+	.phy_id_mask	= 0x0ffffff0,
+	.name		= "Marvell 88E1112 Ethernet",
+	.ops		= &m88e1112_phy_ops,
+};
+
+static struct mii_phy_def *mii_phy_table[] = {
+	&et1011c_phy_def,
+	&cis8201_phy_def,
+	&bcm5248_phy_def,
+	&m88e1111_phy_def,
+	&m88e1112_phy_def,
+	&genmii_phy_def,
+	NULL
+};
+
+int emac_mii_phy_probe(struct mii_phy *phy, int address)
+{
+	struct mii_phy_def *def;
+	int i;
+	u32 id;
+
+	phy->autoneg = AUTONEG_DISABLE;
+	phy->advertising = 0;
+	phy->address = address;
+	phy->speed = SPEED_10;
+	phy->duplex = DUPLEX_HALF;
+	phy->pause = phy->asym_pause = 0;
+
+	/* Take PHY out of isolate mode and reset it. */
+	if (emac_mii_reset_phy(phy))
+		return -ENODEV;
+
+	/* Read ID and find matching entry */
+	id = (phy_read(phy, MII_PHYSID1) << 16) | phy_read(phy, MII_PHYSID2);
+	for (i = 0; (def = mii_phy_table[i]) != NULL; i++)
+		if ((id & def->phy_id_mask) == def->phy_id)
+			break;
+	/* Should never be NULL (we have a generic entry), but... */
+	if (!def)
+		return -ENODEV;
+
+	phy->def = def;
+
+	/* Determine PHY features if needed */
+	phy->features = def->features;
+	if (!phy->features) {
+		u16 bmsr = phy_read(phy, MII_BMSR);
+		if (bmsr & BMSR_ANEGCAPABLE)
+			phy->features |= SUPPORTED_Autoneg;
+		if (bmsr & BMSR_10HALF)
+			phy->features |= SUPPORTED_10baseT_Half;
+		if (bmsr & BMSR_10FULL)
+			phy->features |= SUPPORTED_10baseT_Full;
+		if (bmsr & BMSR_100HALF)
+			phy->features |= SUPPORTED_100baseT_Half;
+		if (bmsr & BMSR_100FULL)
+			phy->features |= SUPPORTED_100baseT_Full;
+		if (bmsr & BMSR_ESTATEN) {
+			u16 esr = phy_read(phy, MII_ESTATUS);
+			if (esr & ESTATUS_1000_TFULL)
+				phy->features |= SUPPORTED_1000baseT_Full;
+			if (esr & ESTATUS_1000_THALF)
+				phy->features |= SUPPORTED_1000baseT_Half;
+		}
+		phy->features |= SUPPORTED_MII;
+	}
+
+	/* Setup default advertising */
+	phy->advertising = phy->features;
+
+	return 0;
+}
+
+MODULE_LICENSE("GPL");
diff --git a/drivers/net/ethernet/ibm/emac/phy.h b/drivers/net/ethernet/ibm/emac/phy.h
new file mode 100644
index 000000000..d7e41ec37
--- /dev/null
+++ b/drivers/net/ethernet/ibm/emac/phy.h
@@ -0,0 +1,87 @@
+/*
+ * drivers/net/ethernet/ibm/emac/phy.h
+ *
+ * Driver for PowerPC 4xx on-chip ethernet controller, PHY support
+ *
+ * Copyright 2007 Benjamin Herrenschmidt, IBM Corp.
+ *                <benh@kernel.crashing.org>
+ *
+ * Based on the arch/ppc version of the driver:
+ *
+ * Benjamin Herrenschmidt <benh@kernel.crashing.org>
+ * February 2003
+ *
+ * Minor additions by Eugene Surovegin <ebs@ebshome.net>, 2004
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ *
+ * This file basically duplicates sungem_phy.{c,h} with different PHYs
+ * supported. I'm looking into merging that in a single mii layer more
+ * flexible than mii.c
+ */
+
+#ifndef __IBM_NEWEMAC_PHY_H
+#define __IBM_NEWEMAC_PHY_H
+
+struct mii_phy;
+
+/* Operations supported by any kind of PHY */
+struct mii_phy_ops {
+	int (*init) (struct mii_phy * phy);
+	int (*suspend) (struct mii_phy * phy, int wol_options);
+	int (*setup_aneg) (struct mii_phy * phy, u32 advertise);
+	int (*setup_forced) (struct mii_phy * phy, int speed, int fd);
+	int (*poll_link) (struct mii_phy * phy);
+	int (*read_link) (struct mii_phy * phy);
+};
+
+/* Structure used to statically define an mii/gii based PHY */
+struct mii_phy_def {
+	u32 phy_id;		/* Concatenated ID1 << 16 | ID2 */
+	u32 phy_id_mask;	/* Significant bits */
+	u32 features;		/* Ethtool SUPPORTED_* defines or
+				   0 for autodetect */
+	int magic_aneg;		/* Autoneg does all speed test for us */
+	const char *name;
+	const struct mii_phy_ops *ops;
+};
+
+/* An instance of a PHY, partially borrowed from mii_if_info */
+struct mii_phy {
+	struct mii_phy_def *def;
+	u32 advertising;	/* Ethtool ADVERTISED_* defines */
+	u32 features;		/* Copied from mii_phy_def.features
+				   or determined automaticaly */
+	int address;		/* PHY address */
+	int mode;		/* PHY mode */
+	int gpcs_address;	/* GPCS PHY address */
+
+	/* 1: autoneg enabled, 0: disabled */
+	int autoneg;
+
+	/* forced speed & duplex (no autoneg)
+	 * partner speed & duplex & pause (autoneg)
+	 */
+	int speed;
+	int duplex;
+	int pause;
+	int asym_pause;
+
+	/* Provided by host chip */
+	struct net_device *dev;
+	int (*mdio_read) (struct net_device * dev, int addr, int reg);
+	void (*mdio_write) (struct net_device * dev, int addr, int reg,
+			    int val);
+};
+
+/* Pass in a struct mii_phy with dev, mdio_read and mdio_write
+ * filled, the remaining fields will be filled on return
+ */
+int emac_mii_phy_probe(struct mii_phy *phy, int address);
+int emac_mii_reset_phy(struct mii_phy *phy);
+int emac_mii_reset_gpcs(struct mii_phy *phy);
+
+#endif /* __IBM_NEWEMAC_PHY_H */
diff --git a/drivers/net/ethernet/ibm/emac/rgmii.c b/drivers/net/ethernet/ibm/emac/rgmii.c
new file mode 100644
index 000000000..206ccbbae
--- /dev/null
+++ b/drivers/net/ethernet/ibm/emac/rgmii.c
@@ -0,0 +1,336 @@
+/*
+ * drivers/net/ethernet/ibm/emac/rgmii.c
+ *
+ * Driver for PowerPC 4xx on-chip ethernet controller, RGMII bridge support.
+ *
+ * Copyright 2007 Benjamin Herrenschmidt, IBM Corp.
+ *                <benh@kernel.crashing.org>
+ *
+ * Based on the arch/ppc version of the driver:
+ *
+ * Copyright (c) 2004, 2005 Zultys Technologies.
+ * Eugene Surovegin <eugene.surovegin@zultys.com> or <ebs@ebshome.net>
+ *
+ * Based on original work by
+ * 	Matt Porter <mporter@kernel.crashing.org>
+ * 	Copyright 2004 MontaVista Software, Inc.
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ *
+ */
+#include <linux/slab.h>
+#include <linux/kernel.h>
+#include <linux/ethtool.h>
+#include <linux/of_address.h>
+#include <asm/io.h>
+
+#include "emac.h"
+#include "debug.h"
+
+// XXX FIXME: Axon seems to support a subset of the RGMII, we
+// thus need to take that into account and possibly change some
+// of the bit settings below that don't seem to quite match the
+// AXON spec
+
+/* RGMIIx_FER */
+#define RGMII_FER_MASK(idx)	(0x7 << ((idx) * 4))
+#define RGMII_FER_RTBI(idx)	(0x4 << ((idx) * 4))
+#define RGMII_FER_RGMII(idx)	(0x5 << ((idx) * 4))
+#define RGMII_FER_TBI(idx)	(0x6 << ((idx) * 4))
+#define RGMII_FER_GMII(idx)	(0x7 << ((idx) * 4))
+#define RGMII_FER_MII(idx)	RGMII_FER_GMII(idx)
+
+/* RGMIIx_SSR */
+#define RGMII_SSR_MASK(idx)	(0x7 << ((idx) * 8))
+#define RGMII_SSR_10(idx)	(0x1 << ((idx) * 8))
+#define RGMII_SSR_100(idx)	(0x2 << ((idx) * 8))
+#define RGMII_SSR_1000(idx)	(0x4 << ((idx) * 8))
+
+/* RGMII bridge supports only GMII/TBI and RGMII/RTBI PHYs */
+static inline int rgmii_valid_mode(int phy_mode)
+{
+	return  phy_mode == PHY_MODE_GMII ||
+		phy_mode == PHY_MODE_MII ||
+		phy_mode == PHY_MODE_RGMII ||
+		phy_mode == PHY_MODE_TBI ||
+		phy_mode == PHY_MODE_RTBI;
+}
+
+static inline const char *rgmii_mode_name(int mode)
+{
+	switch (mode) {
+	case PHY_MODE_RGMII:
+		return "RGMII";
+	case PHY_MODE_TBI:
+		return "TBI";
+	case PHY_MODE_GMII:
+		return "GMII";
+	case PHY_MODE_MII:
+		return "MII";
+	case PHY_MODE_RTBI:
+		return "RTBI";
+	default:
+		BUG();
+	}
+}
+
+static inline u32 rgmii_mode_mask(int mode, int input)
+{
+	switch (mode) {
+	case PHY_MODE_RGMII:
+		return RGMII_FER_RGMII(input);
+	case PHY_MODE_TBI:
+		return RGMII_FER_TBI(input);
+	case PHY_MODE_GMII:
+		return RGMII_FER_GMII(input);
+	case PHY_MODE_MII:
+		return RGMII_FER_MII(input);
+	case PHY_MODE_RTBI:
+		return RGMII_FER_RTBI(input);
+	default:
+		BUG();
+	}
+}
+
+int rgmii_attach(struct platform_device *ofdev, int input, int mode)
+{
+	struct rgmii_instance *dev = platform_get_drvdata(ofdev);
+	struct rgmii_regs __iomem *p = dev->base;
+
+	RGMII_DBG(dev, "attach(%d)" NL, input);
+
+	/* Check if we need to attach to a RGMII */
+	if (input < 0 || !rgmii_valid_mode(mode)) {
+		printk(KERN_ERR "%s: unsupported settings !\n",
+		       ofdev->dev.of_node->full_name);
+		return -ENODEV;
+	}
+
+	mutex_lock(&dev->lock);
+
+	/* Enable this input */
+	out_be32(&p->fer, in_be32(&p->fer) | rgmii_mode_mask(mode, input));
+
+	printk(KERN_NOTICE "%s: input %d in %s mode\n",
+	       ofdev->dev.of_node->full_name, input, rgmii_mode_name(mode));
+
+	++dev->users;
+
+	mutex_unlock(&dev->lock);
+
+	return 0;
+}
+
+void rgmii_set_speed(struct platform_device *ofdev, int input, int speed)
+{
+	struct rgmii_instance *dev = platform_get_drvdata(ofdev);
+	struct rgmii_regs __iomem *p = dev->base;
+	u32 ssr;
+
+	mutex_lock(&dev->lock);
+
+	ssr = in_be32(&p->ssr) & ~RGMII_SSR_MASK(input);
+
+	RGMII_DBG(dev, "speed(%d, %d)" NL, input, speed);
+
+	if (speed == SPEED_1000)
+		ssr |= RGMII_SSR_1000(input);
+	else if (speed == SPEED_100)
+		ssr |= RGMII_SSR_100(input);
+	else if (speed == SPEED_10)
+		ssr |= RGMII_SSR_10(input);
+
+	out_be32(&p->ssr, ssr);
+
+	mutex_unlock(&dev->lock);
+}
+
+void rgmii_get_mdio(struct platform_device *ofdev, int input)
+{
+	struct rgmii_instance *dev = platform_get_drvdata(ofdev);
+	struct rgmii_regs __iomem *p = dev->base;
+	u32 fer;
+
+	RGMII_DBG2(dev, "get_mdio(%d)" NL, input);
+
+	if (!(dev->flags & EMAC_RGMII_FLAG_HAS_MDIO))
+		return;
+
+	mutex_lock(&dev->lock);
+
+	fer = in_be32(&p->fer);
+	fer |= 0x00080000u >> input;
+	out_be32(&p->fer, fer);
+	(void)in_be32(&p->fer);
+
+	DBG2(dev, " fer = 0x%08x\n", fer);
+}
+
+void rgmii_put_mdio(struct platform_device *ofdev, int input)
+{
+	struct rgmii_instance *dev = platform_get_drvdata(ofdev);
+	struct rgmii_regs __iomem *p = dev->base;
+	u32 fer;
+
+	RGMII_DBG2(dev, "put_mdio(%d)" NL, input);
+
+	if (!(dev->flags & EMAC_RGMII_FLAG_HAS_MDIO))
+		return;
+
+	fer = in_be32(&p->fer);
+	fer &= ~(0x00080000u >> input);
+	out_be32(&p->fer, fer);
+	(void)in_be32(&p->fer);
+
+	DBG2(dev, " fer = 0x%08x\n", fer);
+
+	mutex_unlock(&dev->lock);
+}
+
+void rgmii_detach(struct platform_device *ofdev, int input)
+{
+	struct rgmii_instance *dev = platform_get_drvdata(ofdev);
+	struct rgmii_regs __iomem *p;
+
+	BUG_ON(!dev || dev->users == 0);
+	p = dev->base;
+
+	mutex_lock(&dev->lock);
+
+	RGMII_DBG(dev, "detach(%d)" NL, input);
+
+	/* Disable this input */
+	out_be32(&p->fer, in_be32(&p->fer) & ~RGMII_FER_MASK(input));
+
+	--dev->users;
+
+	mutex_unlock(&dev->lock);
+}
+
+int rgmii_get_regs_len(struct platform_device *ofdev)
+{
+	return sizeof(struct emac_ethtool_regs_subhdr) +
+		sizeof(struct rgmii_regs);
+}
+
+void *rgmii_dump_regs(struct platform_device *ofdev, void *buf)
+{
+	struct rgmii_instance *dev = platform_get_drvdata(ofdev);
+	struct emac_ethtool_regs_subhdr *hdr = buf;
+	struct rgmii_regs *regs = (struct rgmii_regs *)(hdr + 1);
+
+	hdr->version = 0;
+	hdr->index = 0; /* for now, are there chips with more than one
+			 * rgmii ? if yes, then we'll add a cell_index
+			 * like we do for emac
+			 */
+	memcpy_fromio(regs, dev->base, sizeof(struct rgmii_regs));
+	return regs + 1;
+}
+
+
+static int rgmii_probe(struct platform_device *ofdev)
+{
+	struct device_node *np = ofdev->dev.of_node;
+	struct rgmii_instance *dev;
+	struct resource regs;
+	int rc;
+
+	rc = -ENOMEM;
+	dev = kzalloc(sizeof(struct rgmii_instance), GFP_KERNEL);
+	if (dev == NULL)
+		goto err_gone;
+
+	mutex_init(&dev->lock);
+	dev->ofdev = ofdev;
+
+	rc = -ENXIO;
+	if (of_address_to_resource(np, 0, &regs)) {
+		printk(KERN_ERR "%s: Can't get registers address\n",
+		       np->full_name);
+		goto err_free;
+	}
+
+	rc = -ENOMEM;
+	dev->base = (struct rgmii_regs __iomem *)ioremap(regs.start,
+						 sizeof(struct rgmii_regs));
+	if (dev->base == NULL) {
+		printk(KERN_ERR "%s: Can't map device registers!\n",
+		       np->full_name);
+		goto err_free;
+	}
+
+	/* Check for RGMII flags */
+	if (of_get_property(ofdev->dev.of_node, "has-mdio", NULL))
+		dev->flags |= EMAC_RGMII_FLAG_HAS_MDIO;
+
+	/* CAB lacks the right properties, fix this up */
+	if (of_device_is_compatible(ofdev->dev.of_node, "ibm,rgmii-axon"))
+		dev->flags |= EMAC_RGMII_FLAG_HAS_MDIO;
+
+	DBG2(dev, " Boot FER = 0x%08x, SSR = 0x%08x\n",
+	     in_be32(&dev->base->fer), in_be32(&dev->base->ssr));
+
+	/* Disable all inputs by default */
+	out_be32(&dev->base->fer, 0);
+
+	printk(KERN_INFO
+	       "RGMII %s initialized with%s MDIO support\n",
+	       ofdev->dev.of_node->full_name,
+	       (dev->flags & EMAC_RGMII_FLAG_HAS_MDIO) ? "" : "out");
+
+	wmb();
+	platform_set_drvdata(ofdev, dev);
+
+	return 0;
+
+ err_free:
+	kfree(dev);
+ err_gone:
+	return rc;
+}
+
+static int rgmii_remove(struct platform_device *ofdev)
+{
+	struct rgmii_instance *dev = platform_get_drvdata(ofdev);
+
+	WARN_ON(dev->users != 0);
+
+	iounmap(dev->base);
+	kfree(dev);
+
+	return 0;
+}
+
+static const struct of_device_id rgmii_match[] =
+{
+	{
+		.compatible	= "ibm,rgmii",
+	},
+	{
+		.type		= "emac-rgmii",
+	},
+	{},
+};
+
+static struct platform_driver rgmii_driver = {
+	.driver = {
+		.name = "emac-rgmii",
+		.of_match_table = rgmii_match,
+	},
+	.probe = rgmii_probe,
+	.remove = rgmii_remove,
+};
+
+int __init rgmii_init(void)
+{
+	return platform_driver_register(&rgmii_driver);
+}
+
+void rgmii_exit(void)
+{
+	platform_driver_unregister(&rgmii_driver);
+}
diff --git a/drivers/net/ethernet/ibm/emac/rgmii.h b/drivers/net/ethernet/ibm/emac/rgmii.h
new file mode 100644
index 000000000..d4f1374d1
--- /dev/null
+++ b/drivers/net/ethernet/ibm/emac/rgmii.h
@@ -0,0 +1,82 @@
+/*
+ * drivers/net/ethernet/ibm/emac/rgmii.h
+ *
+ * Driver for PowerPC 4xx on-chip ethernet controller, RGMII bridge support.
+ *
+ * Copyright 2007 Benjamin Herrenschmidt, IBM Corp.
+ *                <benh@kernel.crashing.org>
+ *
+ * Based on the arch/ppc version of the driver:
+ *
+ * Based on ocp_zmii.h/ibm_emac_zmii.h
+ * Armin Kuster akuster@mvista.com
+ *
+ * Copyright 2004 MontaVista Software, Inc.
+ * Matt Porter <mporter@kernel.crashing.org>
+ *
+ * Copyright (c) 2004, 2005 Zultys Technologies.
+ * Eugene Surovegin <eugene.surovegin@zultys.com> or <ebs@ebshome.net>
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ */
+
+#ifndef __IBM_NEWEMAC_RGMII_H
+#define __IBM_NEWEMAC_RGMII_H
+
+/* RGMII bridge type */
+#define RGMII_STANDARD		0
+#define RGMII_AXON		1
+
+/* RGMII bridge */
+struct rgmii_regs {
+	u32 fer;		/* Function enable register */
+	u32 ssr;		/* Speed select register */
+};
+
+/* RGMII device */
+struct rgmii_instance {
+	struct rgmii_regs __iomem	*base;
+
+	/* RGMII bridge flags */
+	int				flags;
+#define EMAC_RGMII_FLAG_HAS_MDIO	0x00000001
+
+	/* Only one EMAC whacks us at a time */
+	struct mutex			lock;
+
+	/* number of EMACs using this RGMII bridge */
+	int				users;
+
+	/* OF device instance */
+	struct platform_device		*ofdev;
+};
+
+#ifdef CONFIG_IBM_EMAC_RGMII
+
+int rgmii_init(void);
+void rgmii_exit(void);
+int rgmii_attach(struct platform_device *ofdev, int input, int mode);
+void rgmii_detach(struct platform_device *ofdev, int input);
+void rgmii_get_mdio(struct platform_device *ofdev, int input);
+void rgmii_put_mdio(struct platform_device *ofdev, int input);
+void rgmii_set_speed(struct platform_device *ofdev, int input, int speed);
+int rgmii_get_regs_len(struct platform_device *ofdev);
+void *rgmii_dump_regs(struct platform_device *ofdev, void *buf);
+
+#else
+
+# define rgmii_init()		0
+# define rgmii_exit()		do { } while(0)
+# define rgmii_attach(x,y,z)	(-ENXIO)
+# define rgmii_detach(x,y)	do { } while(0)
+# define rgmii_get_mdio(o,i)	do { } while (0)
+# define rgmii_put_mdio(o,i)	do { } while (0)
+# define rgmii_set_speed(x,y,z)	do { } while(0)
+# define rgmii_get_regs_len(x)	0
+# define rgmii_dump_regs(x,buf)	(buf)
+#endif				/* !CONFIG_IBM_EMAC_RGMII */
+
+#endif /* __IBM_NEWEMAC_RGMII_H */
diff --git a/drivers/net/ethernet/ibm/emac/tah.c b/drivers/net/ethernet/ibm/emac/tah.c
new file mode 100644
index 000000000..32cb6c900
--- /dev/null
+++ b/drivers/net/ethernet/ibm/emac/tah.c
@@ -0,0 +1,180 @@
+/*
+ * drivers/net/ethernet/ibm/emac/tah.c
+ *
+ * Driver for PowerPC 4xx on-chip ethernet controller, TAH support.
+ *
+ * Copyright 2007 Benjamin Herrenschmidt, IBM Corp.
+ *                <benh@kernel.crashing.org>
+ *
+ * Based on the arch/ppc version of the driver:
+ *
+ * Copyright 2004 MontaVista Software, Inc.
+ * Matt Porter <mporter@kernel.crashing.org>
+ *
+ * Copyright (c) 2005 Eugene Surovegin <ebs@ebshome.net>
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ */
+#include <linux/of_address.h>
+#include <asm/io.h>
+
+#include "emac.h"
+#include "core.h"
+
+int tah_attach(struct platform_device *ofdev, int channel)
+{
+	struct tah_instance *dev = platform_get_drvdata(ofdev);
+
+	mutex_lock(&dev->lock);
+	/* Reset has been done at probe() time... nothing else to do for now */
+	++dev->users;
+	mutex_unlock(&dev->lock);
+
+	return 0;
+}
+
+void tah_detach(struct platform_device *ofdev, int channel)
+{
+	struct tah_instance *dev = platform_get_drvdata(ofdev);
+
+	mutex_lock(&dev->lock);
+	--dev->users;
+	mutex_unlock(&dev->lock);
+}
+
+void tah_reset(struct platform_device *ofdev)
+{
+	struct tah_instance *dev = platform_get_drvdata(ofdev);
+	struct tah_regs __iomem *p = dev->base;
+	int n;
+
+	/* Reset TAH */
+	out_be32(&p->mr, TAH_MR_SR);
+	n = 100;
+	while ((in_be32(&p->mr) & TAH_MR_SR) && n)
+		--n;
+
+	if (unlikely(!n))
+		printk(KERN_ERR "%s: reset timeout\n",
+			ofdev->dev.of_node->full_name);
+
+	/* 10KB TAH TX FIFO accommodates the max MTU of 9000 */
+	out_be32(&p->mr,
+		 TAH_MR_CVR | TAH_MR_ST_768 | TAH_MR_TFS_10KB | TAH_MR_DTFP |
+		 TAH_MR_DIG);
+}
+
+int tah_get_regs_len(struct platform_device *ofdev)
+{
+	return sizeof(struct emac_ethtool_regs_subhdr) +
+		sizeof(struct tah_regs);
+}
+
+void *tah_dump_regs(struct platform_device *ofdev, void *buf)
+{
+	struct tah_instance *dev = platform_get_drvdata(ofdev);
+	struct emac_ethtool_regs_subhdr *hdr = buf;
+	struct tah_regs *regs = (struct tah_regs *)(hdr + 1);
+
+	hdr->version = 0;
+	hdr->index = 0; /* for now, are there chips with more than one
+			 * zmii ? if yes, then we'll add a cell_index
+			 * like we do for emac
+			 */
+	memcpy_fromio(regs, dev->base, sizeof(struct tah_regs));
+	return regs + 1;
+}
+
+static int tah_probe(struct platform_device *ofdev)
+{
+	struct device_node *np = ofdev->dev.of_node;
+	struct tah_instance *dev;
+	struct resource regs;
+	int rc;
+
+	rc = -ENOMEM;
+	dev = kzalloc(sizeof(struct tah_instance), GFP_KERNEL);
+	if (dev == NULL)
+		goto err_gone;
+
+	mutex_init(&dev->lock);
+	dev->ofdev = ofdev;
+
+	rc = -ENXIO;
+	if (of_address_to_resource(np, 0, &regs)) {
+		printk(KERN_ERR "%s: Can't get registers address\n",
+		       np->full_name);
+		goto err_free;
+	}
+
+	rc = -ENOMEM;
+	dev->base = (struct tah_regs __iomem *)ioremap(regs.start,
+					       sizeof(struct tah_regs));
+	if (dev->base == NULL) {
+		printk(KERN_ERR "%s: Can't map device registers!\n",
+		       np->full_name);
+		goto err_free;
+	}
+
+	platform_set_drvdata(ofdev, dev);
+
+	/* Initialize TAH and enable IPv4 checksum verification, no TSO yet */
+	tah_reset(ofdev);
+
+	printk(KERN_INFO
+	       "TAH %s initialized\n", ofdev->dev.of_node->full_name);
+	wmb();
+
+	return 0;
+
+ err_free:
+	kfree(dev);
+ err_gone:
+	return rc;
+}
+
+static int tah_remove(struct platform_device *ofdev)
+{
+	struct tah_instance *dev = platform_get_drvdata(ofdev);
+
+	WARN_ON(dev->users != 0);
+
+	iounmap(dev->base);
+	kfree(dev);
+
+	return 0;
+}
+
+static const struct of_device_id tah_match[] =
+{
+	{
+		.compatible	= "ibm,tah",
+	},
+	/* For backward compat with old DT */
+	{
+		.type		= "tah",
+	},
+	{},
+};
+
+static struct platform_driver tah_driver = {
+	.driver = {
+		.name = "emac-tah",
+		.of_match_table = tah_match,
+	},
+	.probe = tah_probe,
+	.remove = tah_remove,
+};
+
+int __init tah_init(void)
+{
+	return platform_driver_register(&tah_driver);
+}
+
+void tah_exit(void)
+{
+	platform_driver_unregister(&tah_driver);
+}
diff --git a/drivers/net/ethernet/ibm/emac/tah.h b/drivers/net/ethernet/ibm/emac/tah.h
new file mode 100644
index 000000000..4d5f336f0
--- /dev/null
+++ b/drivers/net/ethernet/ibm/emac/tah.h
@@ -0,0 +1,95 @@
+/*
+ * drivers/net/ethernet/ibm/emac/tah.h
+ *
+ * Driver for PowerPC 4xx on-chip ethernet controller, TAH support.
+ *
+ * Copyright 2007 Benjamin Herrenschmidt, IBM Corp.
+ *                <benh@kernel.crashing.org>
+ *
+ * Based on the arch/ppc version of the driver:
+ *
+ * Copyright 2004 MontaVista Software, Inc.
+ * Matt Porter <mporter@kernel.crashing.org>
+ *
+ * Copyright (c) 2005 Eugene Surovegin <ebs@ebshome.net>
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ */
+
+#ifndef __IBM_NEWEMAC_TAH_H
+#define __IBM_NEWEMAC_TAH_H
+
+/* TAH */
+struct tah_regs {
+	u32 revid;
+	u32 pad[3];
+	u32 mr;
+	u32 ssr0;
+	u32 ssr1;
+	u32 ssr2;
+	u32 ssr3;
+	u32 ssr4;
+	u32 ssr5;
+	u32 tsr;
+};
+
+
+/* TAH device */
+struct tah_instance {
+	struct tah_regs __iomem		*base;
+
+	/* Only one EMAC whacks us at a time */
+	struct mutex			lock;
+
+	/* number of EMACs using this TAH */
+	int				users;
+
+	/* OF device instance */
+	struct platform_device		*ofdev;
+};
+
+
+/* TAH engine */
+#define TAH_MR_CVR		0x80000000
+#define TAH_MR_SR		0x40000000
+#define TAH_MR_ST_256		0x01000000
+#define TAH_MR_ST_512		0x02000000
+#define TAH_MR_ST_768		0x03000000
+#define TAH_MR_ST_1024		0x04000000
+#define TAH_MR_ST_1280		0x05000000
+#define TAH_MR_ST_1536		0x06000000
+#define TAH_MR_TFS_16KB		0x00000000
+#define TAH_MR_TFS_2KB		0x00200000
+#define TAH_MR_TFS_4KB		0x00400000
+#define TAH_MR_TFS_6KB		0x00600000
+#define TAH_MR_TFS_8KB		0x00800000
+#define TAH_MR_TFS_10KB		0x00a00000
+#define TAH_MR_DTFP		0x00100000
+#define TAH_MR_DIG		0x00080000
+
+#ifdef CONFIG_IBM_EMAC_TAH
+
+int tah_init(void);
+void tah_exit(void);
+int tah_attach(struct platform_device *ofdev, int channel);
+void tah_detach(struct platform_device *ofdev, int channel);
+void tah_reset(struct platform_device *ofdev);
+int tah_get_regs_len(struct platform_device *ofdev);
+void *tah_dump_regs(struct platform_device *ofdev, void *buf);
+
+#else
+
+# define tah_init()		0
+# define tah_exit()		do { } while(0)
+# define tah_attach(x,y)	(-ENXIO)
+# define tah_detach(x,y)	do { } while(0)
+# define tah_reset(x)		do { } while(0)
+# define tah_get_regs_len(x)	0
+# define tah_dump_regs(x,buf)	(buf)
+
+#endif				/* !CONFIG_IBM_EMAC_TAH */
+
+#endif /* __IBM_NEWEMAC_TAH_H */
diff --git a/drivers/net/ethernet/ibm/emac/zmii.c b/drivers/net/ethernet/ibm/emac/zmii.c
new file mode 100644
index 000000000..8727b865e
--- /dev/null
+++ b/drivers/net/ethernet/ibm/emac/zmii.c
@@ -0,0 +1,327 @@
+/*
+ * drivers/net/ethernet/ibm/emac/zmii.c
+ *
+ * Driver for PowerPC 4xx on-chip ethernet controller, ZMII bridge support.
+ *
+ * Copyright 2007 Benjamin Herrenschmidt, IBM Corp.
+ *                <benh@kernel.crashing.org>
+ *
+ * Based on the arch/ppc version of the driver:
+ *
+ * Copyright (c) 2004, 2005 Zultys Technologies.
+ * Eugene Surovegin <eugene.surovegin@zultys.com> or <ebs@ebshome.net>
+ *
+ * Based on original work by
+ *      Armin Kuster <akuster@mvista.com>
+ * 	Copyright 2001 MontaVista Softare Inc.
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ *
+ */
+#include <linux/slab.h>
+#include <linux/kernel.h>
+#include <linux/ethtool.h>
+#include <linux/of_address.h>
+#include <asm/io.h>
+
+#include "emac.h"
+#include "core.h"
+
+/* ZMIIx_FER */
+#define ZMII_FER_MDI(idx)	(0x80000000 >> ((idx) * 4))
+#define ZMII_FER_MDI_ALL	(ZMII_FER_MDI(0) | ZMII_FER_MDI(1) | \
+				 ZMII_FER_MDI(2) | ZMII_FER_MDI(3))
+
+#define ZMII_FER_SMII(idx)	(0x40000000 >> ((idx) * 4))
+#define ZMII_FER_RMII(idx)	(0x20000000 >> ((idx) * 4))
+#define ZMII_FER_MII(idx)	(0x10000000 >> ((idx) * 4))
+
+/* ZMIIx_SSR */
+#define ZMII_SSR_SCI(idx)	(0x40000000 >> ((idx) * 4))
+#define ZMII_SSR_FSS(idx)	(0x20000000 >> ((idx) * 4))
+#define ZMII_SSR_SP(idx)	(0x10000000 >> ((idx) * 4))
+
+/* ZMII only supports MII, RMII and SMII
+ * we also support autodetection for backward compatibility
+ */
+static inline int zmii_valid_mode(int mode)
+{
+	return  mode == PHY_MODE_MII ||
+		mode == PHY_MODE_RMII ||
+		mode == PHY_MODE_SMII ||
+		mode == PHY_MODE_NA;
+}
+
+static inline const char *zmii_mode_name(int mode)
+{
+	switch (mode) {
+	case PHY_MODE_MII:
+		return "MII";
+	case PHY_MODE_RMII:
+		return "RMII";
+	case PHY_MODE_SMII:
+		return "SMII";
+	default:
+		BUG();
+	}
+}
+
+static inline u32 zmii_mode_mask(int mode, int input)
+{
+	switch (mode) {
+	case PHY_MODE_MII:
+		return ZMII_FER_MII(input);
+	case PHY_MODE_RMII:
+		return ZMII_FER_RMII(input);
+	case PHY_MODE_SMII:
+		return ZMII_FER_SMII(input);
+	default:
+		return 0;
+	}
+}
+
+int zmii_attach(struct platform_device *ofdev, int input, int *mode)
+{
+	struct zmii_instance *dev = platform_get_drvdata(ofdev);
+	struct zmii_regs __iomem *p = dev->base;
+
+	ZMII_DBG(dev, "init(%d, %d)" NL, input, *mode);
+
+	if (!zmii_valid_mode(*mode)) {
+		/* Probably an EMAC connected to RGMII,
+		 * but it still may need ZMII for MDIO so
+		 * we don't fail here.
+		 */
+		dev->users++;
+		return 0;
+	}
+
+	mutex_lock(&dev->lock);
+
+	/* Autodetect ZMII mode if not specified.
+	 * This is only for backward compatibility with the old driver.
+	 * Please, always specify PHY mode in your board port to avoid
+	 * any surprises.
+	 */
+	if (dev->mode == PHY_MODE_NA) {
+		if (*mode == PHY_MODE_NA) {
+			u32 r = dev->fer_save;
+
+			ZMII_DBG(dev, "autodetecting mode, FER = 0x%08x" NL, r);
+
+			if (r & (ZMII_FER_MII(0) | ZMII_FER_MII(1)))
+				dev->mode = PHY_MODE_MII;
+			else if (r & (ZMII_FER_RMII(0) | ZMII_FER_RMII(1)))
+				dev->mode = PHY_MODE_RMII;
+			else
+				dev->mode = PHY_MODE_SMII;
+		} else
+			dev->mode = *mode;
+
+		printk(KERN_NOTICE "%s: bridge in %s mode\n",
+		       ofdev->dev.of_node->full_name,
+		       zmii_mode_name(dev->mode));
+	} else {
+		/* All inputs must use the same mode */
+		if (*mode != PHY_MODE_NA && *mode != dev->mode) {
+			printk(KERN_ERR
+			       "%s: invalid mode %d specified for input %d\n",
+			       ofdev->dev.of_node->full_name, *mode, input);
+			mutex_unlock(&dev->lock);
+			return -EINVAL;
+		}
+	}
+
+	/* Report back correct PHY mode,
+	 * it may be used during PHY initialization.
+	 */
+	*mode = dev->mode;
+
+	/* Enable this input */
+	out_be32(&p->fer, in_be32(&p->fer) | zmii_mode_mask(dev->mode, input));
+	++dev->users;
+
+	mutex_unlock(&dev->lock);
+
+	return 0;
+}
+
+void zmii_get_mdio(struct platform_device *ofdev, int input)
+{
+	struct zmii_instance *dev = platform_get_drvdata(ofdev);
+	u32 fer;
+
+	ZMII_DBG2(dev, "get_mdio(%d)" NL, input);
+
+	mutex_lock(&dev->lock);
+
+	fer = in_be32(&dev->base->fer) & ~ZMII_FER_MDI_ALL;
+	out_be32(&dev->base->fer, fer | ZMII_FER_MDI(input));
+}
+
+void zmii_put_mdio(struct platform_device *ofdev, int input)
+{
+	struct zmii_instance *dev = platform_get_drvdata(ofdev);
+
+	ZMII_DBG2(dev, "put_mdio(%d)" NL, input);
+	mutex_unlock(&dev->lock);
+}
+
+
+void zmii_set_speed(struct platform_device *ofdev, int input, int speed)
+{
+	struct zmii_instance *dev = platform_get_drvdata(ofdev);
+	u32 ssr;
+
+	mutex_lock(&dev->lock);
+
+	ssr = in_be32(&dev->base->ssr);
+
+	ZMII_DBG(dev, "speed(%d, %d)" NL, input, speed);
+
+	if (speed == SPEED_100)
+		ssr |= ZMII_SSR_SP(input);
+	else
+		ssr &= ~ZMII_SSR_SP(input);
+
+	out_be32(&dev->base->ssr, ssr);
+
+	mutex_unlock(&dev->lock);
+}
+
+void zmii_detach(struct platform_device *ofdev, int input)
+{
+	struct zmii_instance *dev = platform_get_drvdata(ofdev);
+
+	BUG_ON(!dev || dev->users == 0);
+
+	mutex_lock(&dev->lock);
+
+	ZMII_DBG(dev, "detach(%d)" NL, input);
+
+	/* Disable this input */
+	out_be32(&dev->base->fer,
+		 in_be32(&dev->base->fer) & ~zmii_mode_mask(dev->mode, input));
+
+	--dev->users;
+
+	mutex_unlock(&dev->lock);
+}
+
+int zmii_get_regs_len(struct platform_device *ofdev)
+{
+	return sizeof(struct emac_ethtool_regs_subhdr) +
+		sizeof(struct zmii_regs);
+}
+
+void *zmii_dump_regs(struct platform_device *ofdev, void *buf)
+{
+	struct zmii_instance *dev = platform_get_drvdata(ofdev);
+	struct emac_ethtool_regs_subhdr *hdr = buf;
+	struct zmii_regs *regs = (struct zmii_regs *)(hdr + 1);
+
+	hdr->version = 0;
+	hdr->index = 0; /* for now, are there chips with more than one
+			 * zmii ? if yes, then we'll add a cell_index
+			 * like we do for emac
+			 */
+	memcpy_fromio(regs, dev->base, sizeof(struct zmii_regs));
+	return regs + 1;
+}
+
+static int zmii_probe(struct platform_device *ofdev)
+{
+	struct device_node *np = ofdev->dev.of_node;
+	struct zmii_instance *dev;
+	struct resource regs;
+	int rc;
+
+	rc = -ENOMEM;
+	dev = kzalloc(sizeof(struct zmii_instance), GFP_KERNEL);
+	if (dev == NULL)
+		goto err_gone;
+
+	mutex_init(&dev->lock);
+	dev->ofdev = ofdev;
+	dev->mode = PHY_MODE_NA;
+
+	rc = -ENXIO;
+	if (of_address_to_resource(np, 0, &regs)) {
+		printk(KERN_ERR "%s: Can't get registers address\n",
+		       np->full_name);
+		goto err_free;
+	}
+
+	rc = -ENOMEM;
+	dev->base = (struct zmii_regs __iomem *)ioremap(regs.start,
+						sizeof(struct zmii_regs));
+	if (dev->base == NULL) {
+		printk(KERN_ERR "%s: Can't map device registers!\n",
+		       np->full_name);
+		goto err_free;
+	}
+
+	/* We may need FER value for autodetection later */
+	dev->fer_save = in_be32(&dev->base->fer);
+
+	/* Disable all inputs by default */
+	out_be32(&dev->base->fer, 0);
+
+	printk(KERN_INFO
+	       "ZMII %s initialized\n", ofdev->dev.of_node->full_name);
+	wmb();
+	platform_set_drvdata(ofdev, dev);
+
+	return 0;
+
+ err_free:
+	kfree(dev);
+ err_gone:
+	return rc;
+}
+
+static int zmii_remove(struct platform_device *ofdev)
+{
+	struct zmii_instance *dev = platform_get_drvdata(ofdev);
+
+	WARN_ON(dev->users != 0);
+
+	iounmap(dev->base);
+	kfree(dev);
+
+	return 0;
+}
+
+static const struct of_device_id zmii_match[] =
+{
+	{
+		.compatible	= "ibm,zmii",
+	},
+	/* For backward compat with old DT */
+	{
+		.type		= "emac-zmii",
+	},
+	{},
+};
+
+static struct platform_driver zmii_driver = {
+	.driver = {
+		.name = "emac-zmii",
+		.of_match_table = zmii_match,
+	},
+	.probe = zmii_probe,
+	.remove = zmii_remove,
+};
+
+int __init zmii_init(void)
+{
+	return platform_driver_register(&zmii_driver);
+}
+
+void zmii_exit(void)
+{
+	platform_driver_unregister(&zmii_driver);
+}
diff --git a/drivers/net/ethernet/ibm/emac/zmii.h b/drivers/net/ethernet/ibm/emac/zmii.h
new file mode 100644
index 000000000..0959c55b1
--- /dev/null
+++ b/drivers/net/ethernet/ibm/emac/zmii.h
@@ -0,0 +1,78 @@
+/*
+ * drivers/net/ethernet/ibm/emac/zmii.h
+ *
+ * Driver for PowerPC 4xx on-chip ethernet controller, ZMII bridge support.
+ *
+ * Copyright 2007 Benjamin Herrenschmidt, IBM Corp.
+ *                <benh@kernel.crashing.org>
+ *
+ * Based on the arch/ppc version of the driver:
+ *
+ * Copyright (c) 2004, 2005 Zultys Technologies.
+ * Eugene Surovegin <eugene.surovegin@zultys.com> or <ebs@ebshome.net>
+ *
+ * Based on original work by
+ *      Armin Kuster <akuster@mvista.com>
+ * 	Copyright 2001 MontaVista Softare Inc.
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ *
+ */
+#ifndef __IBM_NEWEMAC_ZMII_H
+#define __IBM_NEWEMAC_ZMII_H
+
+/* ZMII bridge registers */
+struct zmii_regs {
+	u32 fer;		/* Function enable reg */
+	u32 ssr;		/* Speed select reg */
+	u32 smiirs;		/* SMII status reg */
+};
+
+/* ZMII device */
+struct zmii_instance {
+	struct zmii_regs __iomem	*base;
+
+	/* Only one EMAC whacks us at a time */
+	struct mutex			lock;
+
+	/* subset of PHY_MODE_XXXX */
+	int				mode;
+
+	/* number of EMACs using this ZMII bridge */
+	int				users;
+
+	/* FER value left by firmware */
+	u32				fer_save;
+
+	/* OF device instance */
+	struct platform_device		*ofdev;
+};
+
+#ifdef CONFIG_IBM_EMAC_ZMII
+
+int zmii_init(void);
+void zmii_exit(void);
+int zmii_attach(struct platform_device *ofdev, int input, int *mode);
+void zmii_detach(struct platform_device *ofdev, int input);
+void zmii_get_mdio(struct platform_device *ofdev, int input);
+void zmii_put_mdio(struct platform_device *ofdev, int input);
+void zmii_set_speed(struct platform_device *ofdev, int input, int speed);
+int zmii_get_regs_len(struct platform_device *ocpdev);
+void *zmii_dump_regs(struct platform_device *ofdev, void *buf);
+
+#else
+# define zmii_init()		0
+# define zmii_exit()		do { } while(0)
+# define zmii_attach(x,y,z)	(-ENXIO)
+# define zmii_detach(x,y)	do { } while(0)
+# define zmii_get_mdio(x,y)	do { } while(0)
+# define zmii_put_mdio(x,y)	do { } while(0)
+# define zmii_set_speed(x,y,z)	do { } while(0)
+# define zmii_get_regs_len(x)	0
+# define zmii_dump_regs(x,buf)	(buf)
+#endif				/* !CONFIG_IBM_EMAC_ZMII */
+
+#endif /* __IBM_NEWEMAC_ZMII_H */