This is a preliminary posting of my Sun Neptune driver against the
net-2.6.24 tree.

The TODO list is very long, but it's a lot of small things,
most of the basic substance of the driver is all there.

Things which are known to be broken:

1) Niagara-2 frontend not written, I have recently received
   the hardware necessary to code that part up.

2) Jumbo MTU.  I have to add some code to adjust the RBR sizes
   when the MTU goes above 1500, should be about an afternoon of
   work.

3) Need to add an API for adding TCAM entries and modifying other
   state.  However, the driver does have full multicast and
   alternate MAC support.

4) Need an API for adding user classification entries, and also
   modifying the flow hashing selections.

I anticipate that #3 and #4 will use a new ethtool interface of
some sort, and this was suggested by Jeff.

Things I know I will change:

1) All the "#if 1" code blocks will disappear or be changed into
   controllable logging statements.  They are there now to catch
   non-fastpath interrupt events etc. for initial driver debugging.

2) All the "#if 0" code blocks will either have users added or
   be removed.

Things which I know are not tested:

1) BMAC and bcm5464r support

I simply lack the hardware which can make use of the 10/100Mbit PHY
support.

The locking also needs to be audited, and I plan to add a lockless
interrupt quiesce scheme like what we use in the tg3 driver.  And
on the topic of interrupts I need to investigate if it's too rude
to try and use so many MSI-X vectors for a single device and in
particular if that can consume all of the MSIs available on some
machines.  Currently the driver just tries to acquire as many as
it can.

Finally, I need to performance tune this driver.  In particular I
am setting the MARK bit on every TX descriptor.  I have to do this
because the chip provides no way to interrupt the host when the
TX queue drains completely.  So if we get some packets before we
hit the threshold for setting the MARK bit, and then get no future
packets, those packet buffers won't be released which is totally
illegal because those buffers are charged against sockets which
cannot be released until the packet buffers get freed.

My initial idea is to use some kind of timer when we have packets
pending in the TX queue but no MARK bits set yet.  As soon as we
set the first MARK bit, we kill the timer.  Otherwise the timer
fires and we process the TX completion queue.

On the performance side there is also the issue of how much I
should pre-pull from the page vector data into the linear packet
buffer.  Basically this is amortizing all the pskb_may_pull()
calls the stack is going to do anyways.  It might be possible to
check out the classification result to figure out how much would
be an optimal pre-pull.

We could also sprinkle some prefetches around the RX path.

The NAPI stuff could be significantly improved especially when we
know that the MSI-X points only to RX DMA events of a specific
channel.  We could avoid all the silly tests on data structures
and also we'd only have to read the V0 and V1 LDG status values.

Anyways, enough yapping, I should save my fingers for implementing
all of the stuff above :-)  This posting is just so people can take
an initial look at the driver.

commit 123f82e3a952726b28c8fb0d6c10d6100e56ff0e
Author: David S. Miller <davem@sunset.davemloft.net>
Date:   Tue Sep 18 14:58:50 2007 -0700

    [NIU]: Add Sun Neptune ethernet driver.
    
    Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/drivers/net/Kconfig b/drivers/net/Kconfig
index fd284a9..8a1d21f 100644
--- a/drivers/net/Kconfig
+++ b/drivers/net/Kconfig
@@ -2638,6 +2638,13 @@ config NETXEN_NIC
 	help
 	  This enables the support for NetXen's Gigabit Ethernet card.
 
+config NIU
+	tristate "Sun Neptune 10Gbit Ethernet support"
+	depends on PCI
+	help
+	  This enables support for cards based upon Sun's
+	  Neptune chipset.
+
 config PASEMI_MAC
 	tristate "PA Semi 1/10Gbit MAC"
 	depends on PPC64 && PCI
diff --git a/drivers/net/Makefile b/drivers/net/Makefile
index 2ab33e8..2706d4d 100644
--- a/drivers/net/Makefile
+++ b/drivers/net/Makefile
@@ -241,3 +241,4 @@ obj-$(CONFIG_NETCONSOLE) += netconsole.o
 obj-$(CONFIG_FS_ENET) += fs_enet/
 
 obj-$(CONFIG_NETXEN_NIC) += netxen/
+obj-$(CONFIG_NIU) += niu.o
diff --git a/drivers/net/niu.c b/drivers/net/niu.c
new file mode 100644
index 0000000..6dfcfac
--- /dev/null
+++ b/drivers/net/niu.c
@@ -0,0 +1,7406 @@
+/* niu.c: Neptune ethernet driver.
+ *
+ * Copyright (C) 2007 David S. Miller (davem@davemloft.net)
+ */
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/pci.h>
+#include <linux/dma-mapping.h>
+#include <linux/netdevice.h>
+#include <linux/ethtool.h>
+#include <linux/etherdevice.h>
+#include <linux/platform_device.h>
+#include <linux/delay.h>
+#include <linux/bitops.h>
+#include <linux/mii.h>
+#include <linux/if_ether.h>
+#include <linux/if_vlan.h>
+#include <linux/ip.h>
+#include <linux/in.h>
+#include <linux/ipv6.h>
+#include <linux/log2.h>
+#include <linux/jiffies.h>
+#include <linux/crc32.h>
+
+#include <net/flow.h>
+
+#include "niu.h"
+
+#define DRV_MODULE_NAME		"niu"
+#define PFX DRV_MODULE_NAME	": "
+#define DRV_MODULE_VERSION	"0.06"
+#define DRV_MODULE_RELDATE	"September 18, 2007"
+
+static char version[] __devinitdata =
+	DRV_MODULE_NAME ".c:v" DRV_MODULE_VERSION " (" DRV_MODULE_RELDATE ")\n";
+
+MODULE_AUTHOR("David S. Miller (davem@davemloft.net)");
+MODULE_DESCRIPTION("NIU ethernet driver");
+MODULE_LICENSE("GPL");
+MODULE_VERSION(DRV_MODULE_VERSION);
+
+#ifndef DMA_44BIT_MASK
+#define DMA_44BIT_MASK	0x00000fffffffffffULL
+#endif
+
+#ifndef PCI_DEVICE_ID_SUN_NEPTUNE
+#define PCI_DEVICE_ID_SUN_NEPTUNE	0xabcd
+#endif
+
+static struct pci_device_id niu_pci_tbl[] = {
+	{PCI_DEVICE(PCI_VENDOR_ID_SUN, PCI_DEVICE_ID_SUN_NEPTUNE),
+	 .driver_data = 0xff},
+	{}
+};
+
+MODULE_DEVICE_TABLE(pci, niu_pci_tbl);
+
+#define NIU_TX_TIMEOUT			(5 * HZ)
+
+#define nr64(reg)		readq(np->regs + (reg))
+#define nw64(val, reg)		writeq((val), np->regs + (reg))
+
+#define nr64_mac(reg)		readq(np->mac_regs + (reg))
+#define nw64_mac(val, reg)	writeq((val), np->mac_regs + (reg))
+
+#define nr64_ipp(reg)		readq(np->regs + np->ipp_off + (reg))
+#define nw64_ipp(val, reg)	writeq((val), np->regs + np->ipp_off + (reg))
+
+#define nr64_pcs(reg)		readq(np->regs + np->pcs_off + (reg))
+#define nw64_pcs(val, reg)	writeq((val), np->regs + np->pcs_off + (reg))
+
+#define nr64_xpcs(reg)		readq(np->regs + np->xpcs_off + (reg))
+#define nw64_xpcs(val, reg)	writeq((val), np->regs + np->xpcs_off + (reg))
+
+static unsigned int niu_debug;
+#define NIU_DEBUG_INTERRUPT	0x00000001
+#define NIU_DEBUG_TX_WORK	0x00000002
+#define NIU_DEBUG_RX_WORK	0x00000004
+#define NIU_DEBUG_POLL		0x00000008
+#define NIU_DEBUG_PROBE		0x00010000
+#define NIU_DEBUG_MDIO		0x00020000
+#define NIU_DEBUG_MII		0x00040000
+#define NIU_DEBUG_INIT_HW	0x00080000
+#define NIU_DEBUG_STOP_HW	0x00080000
+
+module_param(niu_debug, int, 0);
+MODULE_PARM_DESC(niu_debug,
+"NIU bitmapped debugging message enable value:\n"
+"	0x00000001	Log interrupt events\n"
+"	0x00000002	Log TX work\n"
+"	0x00000004	Log RX work\n"
+"	0x00000008	Log NAPI poll\n"
+"	0x00010000	Log device probe events\n"
+"	0x00020000	Log MDIO reads and writes\n"
+"	0x00040000	Log MII reads and writes\n"
+"	0x00080000	Log HW initialization\n"
+"	0x00100000	Log HW shutdown\n"
+);
+
+#define niudbg(TYPE, f, a...) \
+do {	if (niu_debug & NIU_DEBUG_##TYPE) \
+		printk(KERN_ERR PFX f, ## a); \
+} while (0)
+
+#define niu_lock_parent(np, flags) \
+	spin_lock_irqsave(&np->parent->lock, flags)
+#define niu_unlock_parent(np, flags) \
+	spin_unlock_irqrestore(&np->parent->lock, flags)
+
+static int niu_wait_bits_clear_mac(struct niu *np, unsigned long reg, u64 bits,
+				   int limit, int delay)
+{
+	BUILD_BUG_ON(limit <= 0 || delay < 0);
+	while (--limit >= 0) {
+		u64 val = nr64_mac(reg);
+
+		if (!(val & bits))
+			break;
+		udelay(delay);
+	}
+	if (limit < 0)
+		return -ENODEV;
+	return 0;
+}
+
+static int niu_set_and_wait_clear_mac(struct niu *np, unsigned long reg,
+				      u64 bits, int limit, int delay,
+				      const char *reg_name)
+{
+	int err;
+
+	nw64_mac(bits, reg);
+	err = niu_wait_bits_clear_mac(np, reg, bits, limit, delay);
+	if (err)
+		printk(KERN_ERR PFX "%s: bits (%lx) of register %s "
+		       "would not clear, val[%lx]\n",
+		       np->dev->name, bits, reg_name,
+		       nr64_mac(reg));
+	return err;
+}
+
+static int niu_wait_bits_clear_ipp(struct niu *np, unsigned long reg, u64 bits,
+				   int limit, int delay)
+{
+	BUILD_BUG_ON(limit <= 0 || delay < 0);
+	while (--limit >= 0) {
+		u64 val = nr64_ipp(reg);
+
+		if (!(val & bits))
+			break;
+		udelay(delay);
+	}
+	if (limit < 0)
+		return -ENODEV;
+	return 0;
+}
+
+static int niu_set_and_wait_clear_ipp(struct niu *np, unsigned long reg,
+				      u64 bits, int limit, int delay,
+				      const char *reg_name)
+{
+	int err;
+	u64 val;
+
+	val = nr64_ipp(reg);
+	val |= bits;
+	nw64_ipp(val, reg);
+
+	err = niu_wait_bits_clear_ipp(np, reg, bits, limit, delay);
+	if (err)
+		printk(KERN_ERR PFX "%s: bits (%lx) of register %s "
+		       "would not clear, val[%lx]\n",
+		       np->dev->name, bits, reg_name,
+		       nr64_ipp(reg));
+	return err;
+}
+
+static int niu_wait_bits_clear(struct niu *np, unsigned long reg, u64 bits,
+			       int limit, int delay)
+{
+	BUILD_BUG_ON(limit <= 0 || delay < 0);
+	while (--limit >= 0) {
+		u64 val = nr64(reg);
+
+		if (!(val & bits))
+			break;
+		udelay(delay);
+	}
+	if (limit < 0)
+		return -ENODEV;
+	return 0;
+}
+
+static int niu_set_and_wait_clear(struct niu *np, unsigned long reg,
+				  u64 bits, int limit, int delay,
+				  const char *reg_name)
+{
+	int err;
+
+	nw64(bits, reg);
+	err = niu_wait_bits_clear(np, reg, bits, limit, delay);
+	if (err)
+		printk(KERN_ERR PFX "%s: bits (%lx) of register %s "
+		       "would not clear, val[%lx]\n",
+		       np->dev->name, bits, reg_name,
+		       nr64(reg));
+	return err;
+}
+
+static void niu_ldg_rearm(struct niu *np, struct niu_ldg *lp, int on)
+{
+	u64 val = (u64) lp->timer;
+
+	if (on)
+		val |= LDG_IMGMT_ARM;
+
+	nw64(val, LDG_IMGMT(lp->ldg_num));
+}
+
+static int niu_ldn_irq_enable(struct niu *np, int ldn, int on)
+{
+	unsigned long mask_reg, bits;
+	u64 val;
+
+	if (ldn < 0 || ldn > LDN_MAX)
+		return -EINVAL;
+
+	if (ldn < 64) {
+		mask_reg = LD_IM0(ldn);
+		bits = LD_IM0_MASK;
+	} else {
+		mask_reg = LD_IM1(ldn - 64);
+		bits = LD_IM1_MASK;
+	}
+
+	val = nr64(mask_reg);
+	if (on)
+		val &= ~bits;
+	else
+		val |= bits;
+	nw64(val, mask_reg);
+
+	return 0;
+}
+
+static int niu_enable_ldn_in_ldg(struct niu *np, struct niu_ldg *lp, int on)
+{
+	struct niu_parent *parent = np->parent;
+	int i;
+
+	for (i = 0; i <= LDN_MAX; i++) {
+		int err;
+
+		if (parent->ldg_map[i] != lp->ldg_num)
+			continue;
+
+		err = niu_ldn_irq_enable(np, i, on);
+		if (err)
+			return err;
+	}
+	return 0;
+}
+
+static int niu_enable_interrupts(struct niu *np, int on)
+{
+	int i;
+
+	for (i = 0; i < np->num_ldg; i++) {
+		struct niu_ldg *lp = &np->ldg[i];
+		int err;
+
+		err = niu_enable_ldn_in_ldg(np, lp, on);
+		if (err)
+			return err;
+	}
+	for (i = 0; i < np->num_ldg; i++)
+		niu_ldg_rearm(np, &np->ldg[i], on);
+
+	return 0;
+}
+
+static __u32 phy_encode(__u32 type, int port)
+{
+	return (type << (port * 2));
+}
+
+static __u32 phy_decode(__u32 val, int port)
+{
+	return (val >> (port * 2)) & PORT_TYPE_MASK;
+}
+
+static int mdio_wait(struct niu *np)
+{
+	int limit = 1000;
+	u64 val;
+
+	while (--limit > 0) {
+		val = nr64(MIF_FRAME_OUTPUT);
+		if ((val >> MIF_FRAME_OUTPUT_TA_SHIFT) & 0x1)
+			return val & MIF_FRAME_OUTPUT_DATA;
+
+		udelay(10);
+	}
+
+	return -ENODEV;
+}
+
+static int mdio_read(struct niu *np, int port, int dev, int reg)
+{
+	int err;
+
+	nw64(MDIO_ADDR_OP(port, dev, reg), MIF_FRAME_OUTPUT);
+	err = mdio_wait(np);
+	if (err < 0)
+		return err;
+
+	nw64(MDIO_READ_OP(port, dev), MIF_FRAME_OUTPUT);
+	return mdio_wait(np);
+}
+
+static int mdio_write(struct niu *np, int port, int dev, int reg, int data)
+{
+	int err;
+
+	nw64(MDIO_ADDR_OP(port, dev, reg), MIF_FRAME_OUTPUT);
+	err = mdio_wait(np);
+	if (err < 0)
+		return err;
+
+	nw64(MDIO_WRITE_OP(port, dev, data), MIF_FRAME_OUTPUT);
+	err = mdio_wait(np);
+	if (err < 0)
+		return err;
+
+	return 0;
+}
+
+static int mii_read(struct niu *np, int port, int reg)
+{
+	nw64(MII_READ_OP(port, reg), MIF_FRAME_OUTPUT);
+	return mdio_wait(np);
+}
+
+static int mii_write(struct niu *np, int port, int reg, int data)
+{
+	int err;
+
+	nw64(MII_WRITE_OP(port, reg, data), MIF_FRAME_OUTPUT);
+	err = mdio_wait(np);
+	if (err < 0)
+		return err;
+
+	return 0;
+}
+
+static int esr2_set_tx_cfg(struct niu *np, unsigned long channel, u32 val)
+{
+	int err;
+
+	err = mdio_write(np, np->port, NIU_ESR2_DEV_ADDR,
+			 ESR2_TI_PLL_TX_CFG_L(channel),
+			 val & 0xffff);
+	if (!err)
+		err = mdio_write(np, np->port, NIU_ESR2_DEV_ADDR,
+				 ESR2_TI_PLL_TX_CFG_H(channel),
+				 val >> 16);
+	return err;
+}
+
+static int esr2_set_rx_cfg(struct niu *np, unsigned long channel, u32 val)
+{
+	int err;
+
+	err = mdio_write(np, np->port, NIU_ESR2_DEV_ADDR,
+			 ESR2_TI_PLL_RX_CFG_L(channel),
+			 val & 0xffff);
+	if (!err)
+		err = mdio_write(np, np->port, NIU_ESR2_DEV_ADDR,
+				 ESR2_TI_PLL_RX_CFG_H(channel),
+				 val >> 16);
+	return err;
+}
+
+/* Mode is always 10G fiber.  */
+static int serdes_init_niu(struct niu *np)
+{
+	struct niu_link_config *lp = &np->link_config;
+	u32 tx_cfg, rx_cfg;
+	unsigned long i;
+
+	tx_cfg = (PLL_TX_CFG_ENTX | PLL_TX_CFG_SWING_1375MV);
+	rx_cfg = (PLL_RX_CFG_ENRX | PLL_RX_CFG_TERM_0P8VDDT |
+		  PLL_RX_CFG_ALIGN_ENA | PLL_RX_CFG_LOS_LTHRESH |
+		  PLL_RX_CFG_EQ_LP_ADAPTIVE);
+
+	if (lp->loopback_mode == LOOPBACK_PHY) {
+		u16 test_cfg = PLL_TEST_CFG_LOOPBACK_CML_DIS;
+
+		mdio_write(np, np->port, NIU_ESR2_DEV_ADDR,
+			   ESR2_TI_PLL_TEST_CFG_L, test_cfg);
+
+		tx_cfg |= PLL_TX_CFG_ENTEST;
+		rx_cfg |= PLL_RX_CFG_ENTEST;
+	}
+
+	for (i = 0; i < 4; i++) {
+		int err = esr2_set_tx_cfg(np, i, tx_cfg);
+		if (err)
+			return err;
+	}
+
+	for (i = 0; i < 4; i++) {
+		int err = esr2_set_rx_cfg(np, i, rx_cfg);
+		if (err)
+			return err;
+	}
+
+	return 0;
+}
+
+static int esr_read_rxtx_ctrl(struct niu *np, unsigned long chan, u32 *val)
+{
+	int err;
+
+	err = mdio_read(np, np->port, NIU_ESR_DEV_ADDR, ESR_RXTX_CTRL_L(chan));
+	if (err >= 0) {
+		*val = (err & 0xffff);
+		err = mdio_read(np, np->port, NIU_ESR_DEV_ADDR,
+				ESR_RXTX_CTRL_H(chan));
+		if (err >= 0)
+			*val |= ((err & 0xffff) << 16);
+		err = 0;
+	}
+	return err;
+}
+
+static int esr_read_glue0(struct niu *np, unsigned long chan, u32 *val)
+{
+	int err;
+
+	err = mdio_read(np, np->port, NIU_ESR_DEV_ADDR,
+			ESR_GLUE_CTRL0_L(chan));
+	if (err >= 0) {
+		*val = (err & 0xffff);
+		err = mdio_read(np, np->port, NIU_ESR_DEV_ADDR,
+				ESR_GLUE_CTRL0_H(chan));
+		if (err >= 0) {
+			*val |= ((err & 0xffff) << 16);
+			err = 0;
+		}
+	}
+	return err;
+}
+
+static int esr_read_reset(struct niu *np, u32 *val)
+{
+	int err;
+
+	err = mdio_read(np, np->port, NIU_ESR_DEV_ADDR,
+			ESR_RXTX_RESET_CTRL_L);
+	if (err >= 0) {
+		*val = (err & 0xffff);
+		err = mdio_read(np, np->port, NIU_ESR_DEV_ADDR,
+				ESR_RXTX_RESET_CTRL_H);
+		if (err >= 0) {
+			*val |= ((err & 0xffff) << 16);
+			err = 0;
+		}
+	}
+	return err;
+}
+
+static int esr_write_rxtx_ctrl(struct niu *np, unsigned long chan, u32 val)
+{
+	int err;
+
+	err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,
+			 ESR_RXTX_CTRL_L(chan), val & 0xffff);
+	if (!err)
+		err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,
+				 ESR_RXTX_CTRL_H(chan), (val >> 16));
+	return err;
+}
+
+static int esr_write_glue0(struct niu *np, unsigned long chan, u32 val)
+{
+	int err;
+
+	err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,
+			ESR_GLUE_CTRL0_L(chan), val & 0xffff);
+	if (!err)
+		err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,
+				 ESR_GLUE_CTRL0_H(chan), (val >> 16));
+	return err;
+}
+
+static int esr_reset(struct niu *np)
+{
+	u32 reset;
+	int err;
+
+	err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,
+			 ESR_RXTX_RESET_CTRL_L, 0x0000);
+	if (err)
+		return err;
+	err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,
+			 ESR_RXTX_RESET_CTRL_H, 0xffff);
+	if (err)
+		return err;
+	udelay(200);
+
+	
+	err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,
+			 ESR_RXTX_RESET_CTRL_L, 0xffff);
+	if (err)
+		return err;
+	udelay(200);
+
+	err = mdio_write(np, np->port, NIU_ESR_DEV_ADDR,
+			 ESR_RXTX_RESET_CTRL_H, 0x0000);
+	if (err)
+		return err;
+	udelay(200);
+
+	err = esr_read_reset(np, &reset);
+	if (err)
+		return err;
+	if (reset != 0) {
+		printk(KERN_ERR PFX "Port %u ESR_RESET did not clear [%08x]\n",
+		       np->port, reset);
+		return -ENODEV;
+	}
+
+	return 0;
+}
+
+static int serdes_init_10g(struct niu *np)
+{
+	struct niu_link_config *lp = &np->link_config;
+	unsigned long ctrl_reg, test_cfg_reg, i;
+	u64 ctrl_val, test_cfg_val, sig, mask, val;
+	int err;
+
+	switch (np->port) {
+	case 0:
+		ctrl_reg = ENET_SERDES_0_CTRL_CFG;
+		test_cfg_reg = ENET_SERDES_0_TEST_CFG;
+		break;
+	case 1:
+		ctrl_reg = ENET_SERDES_1_CTRL_CFG;
+		test_cfg_reg = ENET_SERDES_1_TEST_CFG;
+		break;
+
+	default:
+		return -EINVAL;
+	}
+	ctrl_val = (ENET_SERDES_CTRL_SDET_0 |
+		    ENET_SERDES_CTRL_SDET_1 |
+		    ENET_SERDES_CTRL_SDET_2 |
+		    ENET_SERDES_CTRL_SDET_3 |
+		    (0x5 << ENET_SERDES_CTRL_EMPH_0_SHIFT) |
+		    (0x5 << ENET_SERDES_CTRL_EMPH_1_SHIFT) |
+		    (0x5 << ENET_SERDES_CTRL_EMPH_2_SHIFT) |
+		    (0x5 << ENET_SERDES_CTRL_EMPH_3_SHIFT) |
+		    (0x1 << ENET_SERDES_CTRL_LADJ_0_SHIFT) |
+		    (0x1 << ENET_SERDES_CTRL_LADJ_1_SHIFT) |
+		    (0x1 << ENET_SERDES_CTRL_LADJ_2_SHIFT) |
+		    (0x1 << ENET_SERDES_CTRL_LADJ_3_SHIFT));
+	test_cfg_val = 0;
+
+	if (lp->loopback_mode == LOOPBACK_PHY) {
+		test_cfg_val |= ((ENET_TEST_MD_PAD_LOOPBACK <<
+				  ENET_SERDES_TEST_MD_0_SHIFT) |
+				 (ENET_TEST_MD_PAD_LOOPBACK <<
+				  ENET_SERDES_TEST_MD_1_SHIFT) |
+				 (ENET_TEST_MD_PAD_LOOPBACK <<
+				  ENET_SERDES_TEST_MD_2_SHIFT) |
+				 (ENET_TEST_MD_PAD_LOOPBACK <<
+				  ENET_SERDES_TEST_MD_3_SHIFT));
+	}
+
+	nw64(ctrl_val, ctrl_reg);
+	nw64(test_cfg_val, test_cfg_reg);
+
+	for (i = 0; i < 4; i++) {
+		u32 rxtx_ctrl, glue0;
+		int err;
+
+		err = esr_read_rxtx_ctrl(np, i, &rxtx_ctrl);
+		if (err)
+			return err;
+		err = esr_read_glue0(np, i, &glue0);
+		if (err)
+			return err;
+
+		rxtx_ctrl &= ~(ESR_RXTX_CTRL_VMUXLO);
+		rxtx_ctrl |= (ESR_RXTX_CTRL_ENSTRETCH |
+			      (2 << ESR_RXTX_CTRL_VMUXLO_SHIFT));
+
+		glue0 &= ~(ESR_GLUE_CTRL0_SRATE |
+			   ESR_GLUE_CTRL0_THCNT |
+			   ESR_GLUE_CTRL0_BLTIME);
+		glue0 |= (ESR_GLUE_CTRL0_RXLOSENAB |
+			  (0xf << ESR_GLUE_CTRL0_SRATE_SHIFT) |
+			  (0xff << ESR_GLUE_CTRL0_THCNT_SHIFT) |
+			  (BLTIME_300_CYCLES <<
+			   ESR_GLUE_CTRL0_BLTIME_SHIFT));
+
+		err = esr_write_rxtx_ctrl(np, i, rxtx_ctrl);
+		if (err)
+			return err;
+		err = esr_write_glue0(np, i, glue0);
+		if (err)
+			return err;
+	}
+
+	err = esr_reset(np);
+	if (err)
+		return err;
+
+	sig = nr64(ESR_INT_SIGNALS);
+	switch (np->port) {
+	case 0:
+		mask = ESR_INT_SIGNALS_P0_BITS;
+		val = (ESR_INT_SRDY0_P0 |
+		       ESR_INT_DET0_P0 |
+		       ESR_INT_XSRDY_P0 |
+		       ESR_INT_XDP_P0_CH3 |
+		       ESR_INT_XDP_P0_CH2 |
+		       ESR_INT_XDP_P0_CH1 |
+		       ESR_INT_XDP_P0_CH0);
+		break;
+
+	case 1:
+		mask = ESR_INT_SIGNALS_P1_BITS;
+		val = (ESR_INT_SRDY0_P1 |
+		       ESR_INT_DET0_P1 |
+		       ESR_INT_XSRDY_P1 |
+		       ESR_INT_XDP_P1_CH3 |
+		       ESR_INT_XDP_P1_CH2 |
+		       ESR_INT_XDP_P1_CH1 |
+		       ESR_INT_XDP_P1_CH0);
+		break;
+
+	default:
+		return -EINVAL;
+	}
+
+	if ((sig & mask) != val) {
+		printk(KERN_ERR PFX "Port %u signal bits [%08lx] are not "
+		       "[%08lx]\n", np->port, (sig & mask), val);
+		return -ENODEV;
+	}
+
+	return 0;
+}
+
+static int serdes_init_1g(struct niu *np)
+{
+	u64 val;
+
+	val = nr64(ENET_SERDES_1_PLL_CFG);
+	val &= ~ENET_SERDES_PLL_FBDIV2;
+	switch (np->port) {
+	case 0:
+		val |= ENET_SERDES_PLL_HRATE0;
+		break;
+	case 1:
+		val |= ENET_SERDES_PLL_HRATE1;
+		break;
+	case 2:
+		val |= ENET_SERDES_PLL_HRATE2;
+		break;
+	case 3:
+		val |= ENET_SERDES_PLL_HRATE3;
+		break;
+	default:
+		return -EINVAL;
+	}
+	nw64(val, ENET_SERDES_1_PLL_CFG);
+
+	return 0;
+}
+
+static int bcm8704_reset(struct niu *np)
+{
+	int err, limit;
+
+	err = mdio_read(np, np->phy_addr,
+			BCM8704_PHYXS_DEV_ADDR, MII_BMCR);
+	if (err < 0)
+		return err;
+	err |= BMCR_RESET;
+	err = mdio_write(np, np->phy_addr, BCM8704_PHYXS_DEV_ADDR,
+			 MII_BMCR, err);
+	if (err)
+		return err;
+
+	limit = 1000;
+	while (--limit >= 0) {
+		err = mdio_read(np, np->phy_addr,
+				BCM8704_PHYXS_DEV_ADDR, MII_BMCR);
+		if (err < 0)
+			return err;
+		if (!(err & BMCR_RESET))
+			break;
+	}
+	if (limit < 0) {
+		printk(KERN_ERR PFX "Port %u PHY will not reset "
+		       "(bmcr=%04x)\n", np->port, (err & 0xffff));
+		return -ENODEV;
+	}
+	return 0;
+}
+
+/* When written, certain PHY registers need to be read back twice
+ * in order for the bits to settle properly.
+ */
+static int bcm8704_user_dev3_readback(struct niu *np, int reg)
+{
+	int err = mdio_read(np, np->phy_addr, BCM8704_USER_DEV3_ADDR, reg);
+	if (err < 0)
+		return err;
+	err = mdio_read(np, np->phy_addr, BCM8704_USER_DEV3_ADDR, reg);
+	if (err < 0)
+		return err;
+	return 0;
+}
+
+static int bcm8704_init_user_dev3(struct niu *np)
+{
+	int err;
+
+	err = mdio_write(np, np->phy_addr,
+			 BCM8704_USER_DEV3_ADDR, BCM8704_USER_CONTROL,
+			 (USER_CONTROL_OPTXRST_LVL |
+			  USER_CONTROL_OPBIASFLT_LVL |
+			  USER_CONTROL_OBTMPFLT_LVL |
+			  USER_CONTROL_OPPRFLT_LVL |
+			  USER_CONTROL_OPTXFLT_LVL |
+			  USER_CONTROL_OPRXLOS_LVL |
+			  USER_CONTROL_OPRXFLT_LVL |
+			  USER_CONTROL_OPTXON_LVL |
+			  (0x3f << USER_CONTROL_RES1_SHIFT)));
+	if (err)
+		return err;
+
+	err = mdio_write(np, np->phy_addr,
+			 BCM8704_USER_DEV3_ADDR, BCM8704_USER_PMD_TX_CONTROL,
+			 (USER_PMD_TX_CTL_XFP_CLKEN |
+			  (1 << USER_PMD_TX_CTL_TX_DAC_TXD_SH) |
+			  (2 << USER_PMD_TX_CTL_TX_DAC_TXCK_SH) |
+			  USER_PMD_TX_CTL_TSCK_LPWREN));
+	if (err)
+		return err;
+
+	err = bcm8704_user_dev3_readback(np, BCM8704_USER_CONTROL);
+	if (err)
+		return err;
+	err = bcm8704_user_dev3_readback(np, BCM8704_USER_PMD_TX_CONTROL);
+	if (err)
+		return err;
+
+	err = mdio_read(np, np->phy_addr, BCM8704_USER_DEV3_ADDR,
+			BCM8704_USER_OPT_DIGITAL_CTRL);
+	if (err < 0)
+		return err;
+	err &= ~USER_ODIG_CTRL_GPIOS;
+	err |= (0x3 << USER_ODIG_CTRL_GPIOS_SHIFT);
+	err = mdio_write(np, np->phy_addr, BCM8704_USER_DEV3_ADDR,
+			 BCM8704_USER_OPT_DIGITAL_CTRL, err);
+	if (err)
+		return err;
+
+	udelay(1000000);
+
+	return 0;
+}
+
+static int xcvr_init_10g(struct niu *np)
+{
+	struct niu_link_config *lp = &np->link_config;
+	u16 analog_stat0, tx_alarm_status;
+	int err;
+	u64 val;
+
+	val = nr64_mac(XMAC_CONFIG);
+	val &= ~XMAC_CONFIG_LED_POLARITY;
+	val |= XMAC_CONFIG_FORCE_LED_ON;
+	nw64_mac(val, XMAC_CONFIG);
+
+	/* XXX shared resource, lock parent XXX */
+	val = nr64(MIF_CONFIG);
+	val |= MIF_CONFIG_INDIRECT_MODE;
+	nw64(val, MIF_CONFIG);
+
+	err = bcm8704_reset(np);
+	if (err)
+		return err;
+
+	err = bcm8704_init_user_dev3(np);
+	if (err)
+		return err;
+
+	err = mdio_read(np, np->phy_addr, BCM8704_PCS_DEV_ADDR,
+			MII_BMCR);
+	if (err < 0)
+		return err;
+	err &= ~BMCR_LOOPBACK;
+
+	if (lp->loopback_mode == LOOPBACK_MAC)
+		err |= BMCR_LOOPBACK;
+
+	err = mdio_write(np, np->phy_addr, BCM8704_PCS_DEV_ADDR,
+			 MII_BMCR, err);
+	if (err)
+		return err;
+
+#if 1
+	err = mdio_read(np, np->phy_addr, BCM8704_PMA_PMD_DEV_ADDR,
+			MII_STAT1000);
+	if (err < 0)
+		return err;
+	printk(KERN_ERR PFX "Port %u PMA_PMD(MII_STAT1000) [%04x]\n",
+	       np->port, err);
+
+	err = mdio_read(np, np->phy_addr, BCM8704_USER_DEV3_ADDR, 0x20);
+	if (err < 0)
+		return err;
+	printk(KERN_ERR PFX "Port %u USER_DEV3(0x20) [%04x]\n",
+	       np->port, err);
+
+	err = mdio_read(np, np->phy_addr, BCM8704_PHYXS_DEV_ADDR,
+			MII_NWAYTEST);
+	if (err < 0)
+		return err;
+	printk(KERN_ERR PFX "Port %u PHYXS(MII_NWAYTEST) [%04x]\n",
+	       np->port, err);
+#endif
+
+	/* XXX dig this out it might not be so useful XXX */
+	err = mdio_read(np, np->phy_addr, BCM8704_USER_DEV3_ADDR,
+			BCM8704_USER_ANALOG_STATUS0);
+	if (err < 0)
+		return err;
+	err = mdio_read(np, np->phy_addr, BCM8704_USER_DEV3_ADDR,
+			BCM8704_USER_ANALOG_STATUS0);
+	if (err < 0)
+		return err;
+	analog_stat0 = err;
+
+	err = mdio_read(np, np->phy_addr, BCM8704_USER_DEV3_ADDR,
+			BCM8704_USER_TX_ALARM_STATUS);
+	if (err < 0)
+		return err;
+	err = mdio_read(np, np->phy_addr, BCM8704_USER_DEV3_ADDR,
+			BCM8704_USER_TX_ALARM_STATUS);
+	if (err < 0)
+		return err;
+	tx_alarm_status = err;
+
+	if (analog_stat0 != 0x03fc) {
+		if ((analog_stat0 == 0x43bc) && (tx_alarm_status != 0)) {
+			printk(KERN_ERR PFX "Port %u cable not connected "
+			       "or bad cable.\n", np->port);
+		} else if (analog_stat0 == 0x639c) {
+			printk(KERN_ERR PFX "Port %u optical module is bad "
+			       "or missing.\n", np->port);
+		}
+	}
+
+	return 0;
+}
+
+static int mii_reset(struct niu *np)
+{
+	int limit, err;
+
+	err = mii_write(np, np->phy_addr, MII_BMCR, BMCR_RESET);
+
+	limit = 1000;
+	while (--limit >= 0) {
+		err = mii_read(np, np->phy_addr, MII_BMCR);
+		if (err < 0)
+			return err;
+		if (!(err & BMCR_RESET))
+			break;
+	}
+	if (limit < 0) {
+		printk(KERN_ERR PFX "Port %u MII would not reset, "
+		       "bmcr[%04x]\n", np->port, err);
+		return -ENODEV;
+	}
+
+	err = mii_write(np, np->phy_addr, MII_BMCR, 0);
+	if (err)
+		return err;
+
+	return 0;
+}
+
+static int mii_init_common(struct niu *np)
+{
+	struct niu_link_config *lp = &np->link_config;
+	u16 bmcr, bmsr, adv, estat;
+	int err;
+
+	err = mii_reset(np);
+	if (err)
+		return err;
+
+	err = mii_read(np, np->phy_addr, MII_BMSR);
+	if (err < 0)
+		return err;
+	bmsr = err;
+
+	estat = 0;
+	if (bmsr & BMSR_ESTATEN) {
+		err = mii_read(np, np->phy_addr, MII_ESTATUS);
+		if (err < 0)
+			return err;
+		estat = err;
+	}
+
+	bmcr = 0;
+
+	if (lp->loopback_mode == LOOPBACK_MAC) {
+		bmcr |= BMCR_LOOPBACK;
+		bmcr &= ~BMCR_ANENABLE;
+		if (lp->active_speed == SPEED_1000)
+			bmcr |= BMCR_SPEED1000;
+		if (lp->active_duplex == DUPLEX_FULL)
+			bmcr |= BMCR_FULLDPLX;
+	} else {
+		bmcr &= ~BMCR_LOOPBACK;
+	}
+	if (lp->loopback_mode == LOOPBACK_PHY) {
+		u16 aux;
+
+		aux = (BCM5464R_AUX_CTL_EXT_LB |
+		       BCM5464R_AUX_CTL_WRITE_1);
+		err = mii_write(np, np->phy_addr, BCM5464R_AUX_CTL, aux);
+		if (err)
+			return err;
+	}
+
+	/* XXX configurable XXX */
+	adv = 0;
+	if (bmsr & BMSR_10HALF)
+		adv |= ADVERTISE_10HALF;
+	if (bmsr & BMSR_10FULL)
+		adv |= ADVERTISE_10FULL;
+	if (bmsr & BMSR_100HALF)
+		adv |= ADVERTISE_100HALF;
+	if (bmsr & BMSR_100FULL)
+		adv |= ADVERTISE_100FULL;
+	if (bmsr & BMSR_100BASE4)
+		adv |= ADVERTISE_100BASE4;
+	adv |= (ADVERTISE_PAUSE_CAP | ADVERTISE_PAUSE_ASYM);
+	err = mii_write(np, np->phy_addr, MII_ADVERTISE, adv);
+	if (err)
+		return err;
+
+	if (bmsr & BMSR_ESTATEN) {
+		u16 ctrl1000;
+
+		ctrl1000 = (BCM5464R_CTRL1000_AS_MASTER |
+			    BCM5464R_CTRL1000_ENABLE_AS_MASTER);
+		if (estat & ESTATUS_1000_TFULL)
+			ctrl1000 |= ADVERTISE_1000FULL;
+		if (estat & ESTATUS_1000_THALF)
+			ctrl1000 |= ADVERTISE_1000HALF;
+		err = mii_write(np, np->phy_addr, MII_CTRL1000, ctrl1000);
+		if (err)
+			return err;
+	}
+	bmcr |= (BMCR_ANENABLE | BMCR_ANRESTART);
+
+	err = mii_write(np, np->phy_addr, MII_BMCR, bmcr);
+	if (err)
+		return err;
+
+	err = mii_read(np, np->phy_addr, MII_BMCR);
+	if (err < 0)
+		return err;
+	err = mii_read(np, np->phy_addr, MII_BMSR);
+	if (err < 0)
+		return err;
+#if 1
+	printk(KERN_ERR PFX "Port %u after MII init bmcr[%04x] bmsr[%04x]\n",
+	       np->port, bmcr, bmsr);
+#endif
+
+	return 0;
+}
+
+static int xcvr_init_1g(struct niu *np)
+{
+	u64 val;
+
+	/* XXX shared resource, lock parent XXX */
+	val = nr64(MIF_CONFIG);
+	val &= ~MIF_CONFIG_INDIRECT_MODE;
+	nw64(val, MIF_CONFIG);
+
+	return mii_init_common(np);
+}
+
+static int niu_xcvr_init(struct niu *np)
+{
+	const struct niu_phy_ops *ops = np->phy_ops;
+	int err;
+
+	err = 0;
+	if (ops->xcvr_init)
+		err = ops->xcvr_init(np);
+
+	return err;
+}
+
+static int niu_serdes_init(struct niu *np)
+{
+	const struct niu_phy_ops *ops = np->phy_ops;
+	int err;
+
+	err = 0;
+	if (ops->serdes_init)
+		err = ops->serdes_init(np);
+
+	return err;
+}
+
+static void niu_init_xif(struct niu *);
+
+static int niu_link_status_common(struct niu *np, int link_up)
+{
+	struct niu_link_config *lp = &np->link_config;
+	struct net_device *dev = np->dev;
+	unsigned long flags;
+
+	if (!netif_carrier_ok(dev) && link_up) {
+		printk(KERN_INFO PFX "%s: Link is up at %s, %s duplex\n",
+		       dev->name,
+		       (lp->active_speed == SPEED_10000 ?
+			"10Gb/sec" :
+			(lp->active_speed == SPEED_1000 ?
+			 "1Gb/sec" :
+			 (lp->active_speed == SPEED_100 ?
+			  "100Mbit/sec" : "10Mbit/sec"))),
+		       (lp->active_duplex == DUPLEX_FULL ?
+			"full" : "half"));
+
+		spin_lock_irqsave(&np->lock, flags);
+		niu_init_xif(np);
+		spin_unlock_irqrestore(&np->lock, flags);
+
+		netif_carrier_on(dev);
+	} else if (netif_carrier_ok(dev) && !link_up) {
+		printk(KERN_WARNING PFX "%s: Link is down\n", dev->name);
+		netif_carrier_off(dev);
+	}
+
+	return 0;
+}
+
+static int link_status_10g(struct niu *np, int *link_up_p)
+{
+	unsigned long flags;
+	int err, link_up;
+
+	if (np->link_config.loopback_mode != LOOPBACK_DISABLED)
+		return -EINVAL;
+
+	link_up = 0;
+
+	spin_lock_irqsave(&np->lock, flags);
+
+	err = mdio_read(np, np->phy_addr, BCM8704_PMA_PMD_DEV_ADDR,
+			BCM8704_PMD_RCV_SIGDET);
+	if (err < 0)
+		return err;
+	if (!(err & PMD_RCV_SIGDET_GLOBAL))
+		goto out;
+
+	err = mdio_read(np, np->phy_addr, BCM8704_PCS_DEV_ADDR,
+			BCM8704_PCS_10G_R_STATUS);
+	if (err < 0)
+		return err;
+	if (!(err & PCS_10G_R_STATUS_BLK_LOCK))
+		goto out;
+
+	err = mdio_read(np, np->phy_addr, BCM8704_PHYXS_DEV_ADDR,
+			BCM8704_PHYXS_XGXS_LANE_STAT);
+	if (err < 0)
+		return err;
+
+	if (err != (PHYXS_XGXS_LANE_STAT_ALINGED |
+		    PHYXS_XGXS_LANE_STAT_MAGIC |
+		    PHYXS_XGXS_LANE_STAT_LANE3 |
+		    PHYXS_XGXS_LANE_STAT_LANE2 |
+		    PHYXS_XGXS_LANE_STAT_LANE1 |
+		    PHYXS_XGXS_LANE_STAT_LANE0))
+		goto out;
+
+	link_up = 1;
+	np->link_config.active_speed = SPEED_10000;
+	np->link_config.active_duplex = DUPLEX_FULL;
+
+out:
+	spin_unlock_irqrestore(&np->lock, flags);
+
+	*link_up_p = link_up;
+	return 0;
+}
+
+static int link_status_1g(struct niu *np, int *link_up_p)
+{
+	u16 current_speed, bmsr;
+	unsigned long flags;
+	u8 current_duplex;
+	int err, link_up;
+
+	if (np->link_config.loopback_mode != LOOPBACK_DISABLED)
+		return -EINVAL;
+
+	link_up = 0;
+	current_speed = SPEED_INVALID;
+	current_duplex = DUPLEX_INVALID;
+
+	spin_lock_irqsave(&np->lock, flags);
+
+	err = mii_read(np, np->phy_addr, MII_BMSR);
+	if (err < 0)
+		goto out;
+
+	bmsr = err;
+	if (bmsr & BMSR_LSTATUS) {
+		u16 adv, lpa, common, estat;
+
+		err = mii_read(np, np->phy_addr, MII_ADVERTISE);
+		if (err < 0)
+			goto out;
+		adv = err;
+
+		err = mii_read(np, np->phy_addr, MII_LPA);
+		if (err < 0)
+			goto out;
+		lpa = err;
+
+		common = adv & lpa;
+
+		err = mii_read(np, np->phy_addr, MII_ESTATUS);
+		if (err < 0)
+			goto out;
+		estat = err;
+
+		if (estat & (ESTATUS_1000_TFULL | ESTATUS_1000_THALF)) {
+			current_speed = SPEED_1000;
+			if (estat == ESTATUS_1000_TFULL)
+				current_duplex = DUPLEX_FULL;
+			else
+				current_duplex = DUPLEX_HALF;
+		} else {
+			if (common & ADVERTISE_100BASE4) {
+				current_speed = SPEED_100;
+				current_duplex = DUPLEX_HALF;
+			} else if (common & ADVERTISE_100FULL) {
+				current_speed = SPEED_100;
+				current_duplex = DUPLEX_FULL;
+			} else if (common & ADVERTISE_100HALF) {
+				current_speed = SPEED_100;
+				current_duplex = DUPLEX_HALF;
+			} else if (common & ADVERTISE_10FULL) {
+				current_speed = SPEED_10;
+				current_duplex = DUPLEX_FULL;
+			} else if (common & ADVERTISE_10HALF) {
+				current_speed = SPEED_10;
+				current_duplex = DUPLEX_HALF;
+			} else
+				goto out;
+		}
+		link_up = 1;
+	}
+
+out:
+	spin_unlock_irqrestore(&np->lock, flags);
+
+	*link_up_p = link_up;
+	return 0;
+}
+
+static int niu_link_status(struct niu *np, int *link_up_p)
+{
+	const struct niu_phy_ops *ops = np->phy_ops;
+	int err;
+
+	err = 0;
+	if (ops->link_status)
+		err = ops->link_status(np, link_up_p);
+
+	return err;
+}
+
+static void niu_timer(unsigned long __opaque)
+{
+	struct niu *np = (struct niu *) __opaque;
+	unsigned long off;
+	int err, link_up;
+
+	err = niu_link_status(np, &link_up);
+	if (!err)
+		niu_link_status_common(np, link_up);
+
+	if (netif_carrier_ok(np->dev))
+		off = 5 * HZ;
+	else
+		off = 1 * HZ;
+	np->timer.expires = jiffies + off;
+
+	add_timer(&np->timer);
+}
+
+static const struct niu_phy_ops phy_ops_10g_fiber_niu = {
+	.serdes_init		= serdes_init_niu,
+	.xcvr_init		= xcvr_init_10g,
+	.link_status		= link_status_10g,
+};
+
+static const struct niu_phy_ops phy_ops_10g_fiber = {
+	.serdes_init		= serdes_init_10g,
+	.xcvr_init		= xcvr_init_10g,
+	.link_status		= link_status_10g,
+};
+
+static const struct niu_phy_ops phy_ops_10g_copper = {
+	.serdes_init		= serdes_init_10g,
+	.link_status		= link_status_10g, /* XXX */
+};
+
+static const struct niu_phy_ops phy_ops_1g_fiber = {
+	.serdes_init		= serdes_init_1g,
+	.xcvr_init		= xcvr_init_1g,
+	.link_status		= link_status_1g,
+};
+
+static const struct niu_phy_ops phy_ops_1g_copper = {
+	.xcvr_init		= xcvr_init_1g,
+	.link_status		= link_status_1g,
+};
+
+struct niu_phy_template {
+	const struct niu_phy_ops	*ops;
+	u32				phy_addr_base;
+};
+
+static const struct niu_phy_template phy_template_niu = {
+	.ops		= &phy_ops_10g_fiber_niu,
+	.phy_addr_base	= 16,
+};
+
+static const struct niu_phy_template phy_template_10g_fiber = {
+	.ops		= &phy_ops_10g_fiber,
+	.phy_addr_base	= 8,
+};
+
+static const struct niu_phy_template phy_template_10g_copper = {
+	.ops		= &phy_ops_10g_copper,
+	.phy_addr_base	= 10,
+};
+
+static const struct niu_phy_template phy_template_1g_fiber = {
+	.ops		= &phy_ops_1g_fiber,
+	.phy_addr_base	= 0,
+};
+
+static const struct niu_phy_template phy_template_1g_copper = {
+	.ops		= &phy_ops_1g_copper,
+	.phy_addr_base	= 0,
+};
+
+static int niu_determine_phy_disposition(struct niu *np)
+{
+	struct niu_parent *parent = np->parent;
+	__u8 plat_type = parent->plat_type;
+	const struct niu_phy_template *tp;
+	__u32 phy_addr_off = 0;
+
+	if (plat_type == PLAT_TYPE_NIU) {
+		tp = &phy_template_niu;
+		phy_addr_off += np->port;
+	} else {
+		switch (np->flags & (NIU_FLAGS_10G | NIU_FLAGS_FIBER)) {
+		case 0:
+			/* 1G copper */
+			tp = &phy_template_1g_copper;
+			if (plat_type == PLAT_TYPE_VF_P0)
+				phy_addr_off = 10;
+			else if (plat_type == PLAT_TYPE_VF_P1)
+				phy_addr_off = 26;
+
+			phy_addr_off += (np->port ^ 0x3);
+			break;
+
+		case NIU_FLAGS_10G:
+			/* 10G copper */
+			tp = &phy_template_1g_copper;
+			break;
+
+		case NIU_FLAGS_FIBER:
+			/* 1G fiber */
+			tp = &phy_template_1g_fiber;
+			break;
+
+		case NIU_FLAGS_10G | NIU_FLAGS_FIBER:
+			/* 10G fiber */
+			tp = &phy_template_10g_fiber;
+			if (plat_type == PLAT_TYPE_VF_P0 ||
+			    plat_type == PLAT_TYPE_VF_P1)
+				phy_addr_off = 8;
+			phy_addr_off += np->port;
+			break;
+
+		default:
+			return -EINVAL;
+		}
+	}
+
+	np->phy_ops = tp->ops;
+	np->phy_addr = tp->phy_addr_base + phy_addr_off;
+
+	return 0;
+}
+
+static int niu_init_link(struct niu *np)
+{
+	struct niu_parent *parent = np->parent;
+	int err, ignore;
+
+	if (parent->plat_type == PLAT_TYPE_NIU) {
+		err = niu_xcvr_init(np);
+		if (err)
+			return err;
+		msleep(200);
+	}
+	err = niu_serdes_init(np);
+	if (err)
+		return err;
+	msleep(200);
+	err = niu_xcvr_init(np);
+	if (!err)
+		niu_link_status(np, &ignore);
+	return 0;
+}
+
+static void niu_set_primary_mac(struct niu *np, unsigned char *addr)
+{
+	u16 reg0 = addr[4] << 8 | addr[5];
+	u16 reg1 = addr[2] << 8 | addr[3];
+	u16 reg2 = addr[0] << 8 | addr[1];
+
+	if (np->flags & NIU_FLAGS_XMAC) {
+		nw64_mac(reg0, XMAC_ADDR0);
+		nw64_mac(reg1, XMAC_ADDR1);
+		nw64_mac(reg2, XMAC_ADDR2);
+	} else {
+		nw64_mac(reg0, BMAC_ADDR0);
+		nw64_mac(reg1, BMAC_ADDR1);
+		nw64_mac(reg2, BMAC_ADDR2);
+	}
+}
+
+static int niu_num_alt_addr(struct niu *np)
+{
+	if (np->flags & NIU_FLAGS_XMAC)
+		return XMAC_NUM_ALT_ADDR;
+	else
+		return BMAC_NUM_ALT_ADDR;
+}
+
+static int niu_set_alt_mac(struct niu *np, int index, unsigned char *addr)
+{
+	u16 reg0 = addr[4] << 8 | addr[5];
+	u16 reg1 = addr[2] << 8 | addr[3];
+	u16 reg2 = addr[0] << 8 | addr[1];
+
+	if (index >= niu_num_alt_addr(np))
+		return -EINVAL;
+
+	if (np->flags & NIU_FLAGS_XMAC) {
+		nw64_mac(reg0, XMAC_ALT_ADDR0(index));
+		nw64_mac(reg1, XMAC_ALT_ADDR1(index));
+		nw64_mac(reg2, XMAC_ALT_ADDR2(index));
+	} else {
+		nw64_mac(reg0, BMAC_ALT_ADDR0(index));
+		nw64_mac(reg1, BMAC_ALT_ADDR1(index));
+		nw64_mac(reg2, BMAC_ALT_ADDR2(index));
+	}
+
+	return 0;
+}
+
+static int niu_enable_alt_mac(struct niu *np, int index, int on)
+{
+	unsigned long reg;
+	u64 val, mask;
+
+	if (index >= niu_num_alt_addr(np))
+		return -EINVAL;
+
+	if (np->flags & NIU_FLAGS_XMAC)
+		reg = XMAC_ADDR_CMPEN;
+	else
+		reg = BMAC_ADDR_CMPEN;
+
+	mask = 1 << index;
+
+	val = nr64_mac(reg);
+	if (on)
+		val |= mask;
+	else
+		val &= ~mask;
+	nw64_mac(val, reg);
+
+	return 0;
+}
+
+static void __set_rdc_table_num_hw(struct niu *np, unsigned long reg,
+				   int num, int mac_pref)
+{
+	u64 val = nr64_mac(reg);
+	val &= ~(HOST_INFO_MACRDCTBLN | HOST_INFO_MPR);
+	val |= num;
+	if (mac_pref)
+		val |= HOST_INFO_MPR;
+	nw64_mac(val, reg);
+}
+
+static int __set_rdc_table_num(struct niu *np,
+			       int xmac_index, int bmac_index,
+			       int rdc_table_num, int mac_pref)
+{
+	unsigned long reg;
+
+	if (rdc_table_num & ~HOST_INFO_MACRDCTBLN)
+		return -EINVAL;
+	if (np->flags & NIU_FLAGS_XMAC)
+		reg = XMAC_HOST_INFO(xmac_index);
+	else
+		reg = BMAC_HOST_INFO(bmac_index);
+	__set_rdc_table_num_hw(np, reg, rdc_table_num, mac_pref);
+	return 0;
+}
+
+static int niu_set_primary_mac_rdc_table(struct niu *np, int table_num,
+					 int mac_pref)
+{
+	return __set_rdc_table_num(np, 17, 0, table_num, mac_pref);
+}
+
+static int niu_set_multicast_mac_rdc_table(struct niu *np, int table_num,
+					   int mac_pref)
+{
+	return __set_rdc_table_num(np, 16, 8, table_num, mac_pref);
+}
+
+static int niu_set_alt_mac_rdc_table(struct niu *np, int idx, int table_num,
+				     int mac_pref)
+{
+	if (idx >= niu_num_alt_addr(np))
+		return -EINVAL;
+	return __set_rdc_table_num(np, idx, idx + 1, table_num, mac_pref);
+}
+
+static u64 vlan_entry_set_parity(u64 reg_val)
+{
+	u64 port01_mask;
+	u64 port23_mask;
+
+	port01_mask = 0x00ff;
+	port23_mask = 0xff00;
+
+	if (hweight64(reg_val & port01_mask) & 1)
+		reg_val |= ENET_VLAN_TBL_PARITY0;
+	else
+		reg_val &= ~ENET_VLAN_TBL_PARITY0;
+
+	if (hweight64(reg_val & port23_mask) & 1)
+		reg_val |= ENET_VLAN_TBL_PARITY1;
+	else
+		reg_val &= ~ENET_VLAN_TBL_PARITY1;
+
+	return reg_val;
+}
+
+static void vlan_tbl_write(struct niu *np, unsigned long index,
+			   int port, int vpr, int rdc_table)
+{
+	u64 reg_val = nr64(ENET_VLAN_TBL(index));
+
+	reg_val &= ~((ENET_VLAN_TBL_VPR |
+		      ENET_VLAN_TBL_VLANRDCTBLN) <<
+		     ENET_VLAN_TBL_SHIFT(port));
+	if (vpr)
+		reg_val |= (ENET_VLAN_TBL_VPR <<
+			    ENET_VLAN_TBL_SHIFT(port));
+	reg_val |= (rdc_table << ENET_VLAN_TBL_SHIFT(port));
+
+	reg_val = vlan_entry_set_parity(reg_val);
+
+	nw64(reg_val, ENET_VLAN_TBL(index));
+}
+
+static void vlan_tbl_clear(struct niu *np)
+{
+	int i;
+
+	for (i = 0; i < ENET_VLAN_TBL_NUM_ENTRIES; i++)
+		nw64(0, ENET_VLAN_TBL(i));
+}
+
+static int tcam_wait_bit(struct niu *np, u64 bit)
+{
+	int limit = 1000;
+
+	while (--limit > 0) {
+		if (nr64(TCAM_CTL) & bit)
+			break;
+		udelay(1);
+	}
+	if (limit < 0)
+		return -ENODEV;
+
+	return 0;
+}
+
+static int tcam_flush(struct niu *np, int index)
+{
+	nw64(0x00, TCAM_KEY_0);
+	nw64(0xff, TCAM_KEY_MASK_0);
+	nw64((TCAM_CTL_RWC_TCAM_WRITE | index), TCAM_CTL);
+
+	return tcam_wait_bit(np, TCAM_CTL_STAT);
+}
+
+#if 0
+static int tcam_read(struct niu *np, int index,
+		     u64 *key, u64 *mask)
+{
+	int err;
+
+	nw64((TCAM_CTL_RWC_TCAM_READ | index), TCAM_CTL);
+	err = tcam_wait_bit(np, TCAM_CTL_STAT);
+	if (!err) {
+		key[0] = nr64(TCAM_KEY_0);
+		key[1] = nr64(TCAM_KEY_1);
+		key[2] = nr64(TCAM_KEY_2);
+		key[3] = nr64(TCAM_KEY_3);
+		mask[0] = nr64(TCAM_KEY_MASK_0);
+		mask[1] = nr64(TCAM_KEY_MASK_1);
+		mask[2] = nr64(TCAM_KEY_MASK_2);
+		mask[3] = nr64(TCAM_KEY_MASK_3);
+	}
+	return err;
+}
+#endif
+
+static int tcam_write(struct niu *np, int index,
+		      u64 *key, u64 *mask)
+{
+	nw64(key[0], TCAM_KEY_0);
+	nw64(key[1], TCAM_KEY_1);
+	nw64(key[2], TCAM_KEY_2);
+	nw64(key[3], TCAM_KEY_3);
+	nw64(mask[0], TCAM_KEY_MASK_0);
+	nw64(mask[1], TCAM_KEY_MASK_1);
+	nw64(mask[2], TCAM_KEY_MASK_2);
+	nw64(mask[3], TCAM_KEY_MASK_3);
+	nw64((TCAM_CTL_RWC_TCAM_WRITE | index), TCAM_CTL);
+
+	return tcam_wait_bit(np, TCAM_CTL_STAT);
+}
+
+#if 0
+static int tcam_assoc_read(struct niu *np, int index, u64 *data)
+{
+	int err;
+
+	nw64((TCAM_CTL_RWC_RAM_READ | index), TCAM_CTL);
+	err = tcam_wait_bit(np, TCAM_CTL_STAT);
+	if (!err)
+		*data = nr64(TCAM_KEY_1);
+
+	return err;
+}
+#endif
+
+static int tcam_assoc_write(struct niu *np, int index, u64 assoc_data)
+{
+	nw64(assoc_data, TCAM_KEY_1);
+	nw64((TCAM_CTL_RWC_RAM_WRITE | index), TCAM_CTL);
+
+	return tcam_wait_bit(np, TCAM_CTL_STAT);
+}
+
+static void tcam_enable(struct niu *np, int on)
+{
+	u64 val = nr64(FFLP_CFG_1);
+
+	if (on)
+		val &= ~FFLP_CFG_1_TCAM_DIS;
+	else
+		val |= FFLP_CFG_1_TCAM_DIS;
+	nw64(val, FFLP_CFG_1);
+}
+
+static void tcam_set_lat_and_ratio(struct niu *np, u64 latency, u64 ratio)
+{
+	u64 val = nr64(FFLP_CFG_1);
+
+	val &= ~(FFLP_CFG_1_FFLPINITDONE |
+		 FFLP_CFG_1_CAMLAT |
+		 FFLP_CFG_1_CAMRATIO);
+	val |= (latency << FFLP_CFG_1_CAMLAT_SHIFT);
+	val |= (ratio << FFLP_CFG_1_CAMRATIO_SHIFT);
+	nw64(val, FFLP_CFG_1);
+
+	val = nr64(FFLP_CFG_1);
+	val |= FFLP_CFG_1_FFLPINITDONE;
+	nw64(val, FFLP_CFG_1);
+}
+
+static int tcam_user_eth_class_enable(struct niu *np, unsigned long class,
+				      int on)
+{
+	unsigned long reg;
+	u64 val;
+
+	if (class < CLASS_CODE_ETHERTYPE1 ||
+	    class > CLASS_CODE_ETHERTYPE2)
+		return -EINVAL;
+
+	reg = L2_CLS(class - CLASS_CODE_ETHERTYPE1);
+	val = nr64(reg);
+	if (on)
+		val |= L2_CLS_VLD;
+	else
+		val &= ~L2_CLS_VLD;
+	nw64(val, reg);
+
+	return 0;
+}
+
+#if 0
+static int tcam_user_eth_class_set(struct niu *np, unsigned long class,
+				   u64 ether_type)
+{
+	unsigned long reg;
+	u64 val;
+
+	if (class < CLASS_CODE_ETHERTYPE1 ||
+	    class > CLASS_CODE_ETHERTYPE2 ||
+	    (ether_type & ~(u64)0xffff) != 0)
+		return -EINVAL;
+
+	reg = L2_CLS(class - CLASS_CODE_ETHERTYPE1);
+	val = nr64(reg);
+	val &= ~L2_CLS_ETYPE;
+	val |= (ether_type << L2_CLS_ETYPE_SHIFT);
+	nw64(val, reg);
+
+	return 0;
+}
+#endif
+
+static int tcam_user_ip_class_enable(struct niu *np, unsigned long class,
+				     int on)
+{
+	unsigned long reg;
+	u64 val;
+
+	if (class < CLASS_CODE_USER_PROG1 ||
+	    class > CLASS_CODE_USER_PROG4)
+		return -EINVAL;
+
+	reg = L3_CLS(class - CLASS_CODE_USER_PROG1);
+	val = nr64(reg);
+	if (on)
+		val |= L3_CLS_VALID;
+	else
+		val &= ~L3_CLS_VALID;
+	nw64(val, reg);
+
+	return 0;
+}
+
+#if 0
+static int tcam_user_ip_class_set(struct niu *np, unsigned long class,
+				  int ipv6, u64 protocol_id,
+				  u64 tos_mask, u64 tos_val)
+{
+	unsigned long reg;
+	u64 val;
+
+	if (class < CLASS_CODE_USER_PROG1 ||
+	    class > CLASS_CODE_USER_PROG4 ||
+	    (protocol_id & ~(u64)0xff) != 0 ||
+	    (tos_mask & ~(u64)0xff) != 0 ||
+	    (tos_val & ~(u64)0xff) != 0)
+		return -EINVAL;
+
+	reg = L3_CLS(class - CLASS_CODE_USER_PROG1);
+	val = nr64(reg);
+	val &= ~(L3_CLS_IPVER | L3_CLS_PID |
+		 L3_CLS_TOSMASK | L3_CLS_TOS);
+	if (ipv6)
+		val |= L3_CLS_IPVER;
+	val |= (protocol_id << L3_CLS_PID_SHIFT);
+	val |= (tos_mask << L3_CLS_TOSMASK_SHIFT);
+	val |= (tos_val << L3_CLS_TOS_SHIFT);
+	nw64(val, reg);
+
+	return 0;
+}
+#endif
+
+static int tcam_early_init(struct niu *np)
+{
+	unsigned long i;
+	int err;
+
+	tcam_enable(np, 0);
+	tcam_set_lat_and_ratio(np,
+			       DEFAULT_TCAM_LATENCY,
+			       DEFAULT_TCAM_ACCESS_RATIO);
+	for (i = CLASS_CODE_ETHERTYPE1; i <= CLASS_CODE_ETHERTYPE2; i++) {
+		err = tcam_user_eth_class_enable(np, i, 0);
+		if (err)
+			return err;
+	}
+	for (i = CLASS_CODE_USER_PROG1; i <= CLASS_CODE_USER_PROG4; i++) {
+		err = tcam_user_ip_class_enable(np, i, 0);
+		if (err)
+			return err;
+	}
+
+	return 0;
+}
+
+static int tcam_flush_all(struct niu *np)
+{
+	unsigned long i;
+
+	for (i = 0; i < np->parent->tcam_num_entries; i++) {
+		int err = tcam_flush(np, i);
+		if (err)
+			return err;
+	}
+	return 0;
+}
+
+static u64 hash_addr_regval(unsigned long index, unsigned long num_entries)
+{
+	return ((u64)index | (num_entries == 1 ?
+			      HASH_TBL_ADDR_AUTOINC : 0));
+}
+
+#if 0
+static int hash_read(struct niu *np, unsigned long partition,
+		     unsigned long index, unsigned long num_entries,
+		     u64 *data)
+{
+	u64 val = hash_addr_regval(index, num_entries);
+	unsigned long i;
+
+	if (partition >= FCRAM_NUM_PARTITIONS ||
+	    index + num_entries > FCRAM_SIZE)
+		return -EINVAL;
+
+	nw64(val, HASH_TBL_ADDR(partition));
+	for (i = 0; i < num_entries; i++)
+		data[i] = nr64(HASH_TBL_DATA(partition));
+
+	return 0;
+}
+#endif
+
+static int hash_write(struct niu *np, unsigned long partition,
+		      unsigned long index, unsigned long num_entries,
+		      u64 *data)
+{
+	u64 val = hash_addr_regval(index, num_entries);
+	unsigned long i;
+
+	if (partition >= FCRAM_NUM_PARTITIONS ||
+	    index + (num_entries * 8) > FCRAM_SIZE)
+		return -EINVAL;
+
+	nw64(val, HASH_TBL_ADDR(partition));
+	for (i = 0; i < num_entries; i++)
+		nw64(data[i], HASH_TBL_DATA(partition));
+
+	return 0;
+}
+		       
+static void fflp_reset(struct niu *np)
+{
+	u64 val;
+
+	nw64(FFLP_CFG_1_PIO_FIO_RST, FFLP_CFG_1);
+	udelay(10);
+	nw64(0, FFLP_CFG_1);
+
+	val = FFLP_CFG_1_FCRAMOUTDR_NORMAL | FFLP_CFG_1_FFLPINITDONE;
+	nw64(val, FFLP_CFG_1);
+}
+
+static void fflp_set_timings(struct niu *np)
+{
+	u64 val = nr64(FFLP_CFG_1);
+
+	val &= ~FFLP_CFG_1_FFLPINITDONE;
+	val |= (DEFAULT_FCRAMRATIO << FFLP_CFG_1_FCRAMRATIO_SHIFT);
+	nw64(val, FFLP_CFG_1);
+
+	val = nr64(FFLP_CFG_1);
+	val |= FFLP_CFG_1_FFLPINITDONE;
+	nw64(val, FFLP_CFG_1);
+
+	val = nr64(FCRAM_REF_TMR);
+	val &= ~(FCRAM_REF_TMR_MAX | FCRAM_REF_TMR_MIN);
+	val |= (DEFAULT_FCRAM_REFRESH_MAX << FCRAM_REF_TMR_MAX_SHIFT);
+	val |= (DEFAULT_FCRAM_REFRESH_MIN << FCRAM_REF_TMR_MIN_SHIFT);
+	nw64(val, FCRAM_REF_TMR);
+}
+
+static int fflp_set_partition(struct niu *np, u64 partition,
+			       u64 mask, u64 base, int enable)
+{
+	unsigned long reg;
+	u64 val;
+
+	if (partition >= FCRAM_NUM_PARTITIONS ||
+	    (mask & ~(u64)0x1f) != 0 ||
+	    (base & ~(u64)0x1f) != 0)
+		return -EINVAL;
+
+	reg = FLW_PRT_SEL(partition);
+
+	val = nr64(reg);
+	val &= ~(FLW_PRT_SEL_EXT | FLW_PRT_SEL_MASK | FLW_PRT_SEL_BASE);
+	val |= (mask << FLW_PRT_SEL_MASK_SHIFT);
+	val |= (base << FLW_PRT_SEL_BASE_SHIFT);
+	if (enable)
+		val |= FLW_PRT_SEL_EXT;
+	nw64(val, reg);
+
+	return 0;
+}
+
+static int fflp_disable_all_partitions(struct niu *np)
+{
+	unsigned long i;
+
+	for (i = 0; i < FCRAM_NUM_PARTITIONS; i++) {
+		int err = fflp_set_partition(np, 0, 0, 0, 0);
+		if (err)
+			return err;
+	}
+	return 0;
+}
+
+static void fflp_llcsnap_enable(struct niu *np, int on)
+{
+	u64 val = nr64(FFLP_CFG_1);
+
+	if (on)
+		val |= FFLP_CFG_1_LLCSNAP;
+	else
+		val &= ~FFLP_CFG_1_LLCSNAP;
+	nw64(val, FFLP_CFG_1);
+}
+
+static void fflp_errors_enable(struct niu *np, int on)
+{
+	u64 val = nr64(FFLP_CFG_1);
+
+	if (on)
+		val &= ~FFLP_CFG_1_ERRORDIS;
+	else
+		val |= FFLP_CFG_1_ERRORDIS;
+	nw64(val, FFLP_CFG_1);
+}
+
+static int fflp_hash_clear(struct niu *np)
+{
+	struct fcram_hash_ipv4 ent;
+	unsigned long i;
+
+	/* IPV4 hash entry with valid bit clear, rest is don't care.  */
+	memset(&ent, 0, sizeof(ent));
+	ent.header = HASH_HEADER_EXT;
+
+	for (i = 0; i < FCRAM_SIZE; i += sizeof(ent)) {
+		int err = hash_write(np, 0, i, 1, (u64 *) &ent);
+		if (err)
+			return err;
+	}
+	return 0;
+}
+
+static int fflp_early_init(struct niu *np)
+{
+	struct niu_parent *parent;
+	unsigned long flags;
+	int err;
+
+	niu_lock_parent(np, flags);
+
+	parent = np->parent;
+	err = 0;
+	if (!(parent->flags & PARENT_FLGS_CLS_HWINIT)) {
+		niudbg(PROBE, "fflp_early_init: Initting hw on port %u\n",
+		       np->port);
+		if (np->parent->plat_type == PLAT_TYPE_ATLAS) {
+			fflp_reset(np);
+			fflp_set_timings(np);
+			err = fflp_disable_all_partitions(np);
+			if (err) {
+				niudbg(PROBE, "fflp_disable_all_partitions "
+				       "failed, err=%d\n", err);
+				goto out;
+			}
+		}
+
+		err = tcam_early_init(np);
+		if (err) {
+			niudbg(PROBE, "tcam_early_init failed, err=%d\n",
+			       err);
+			goto out;
+		}
+		fflp_llcsnap_enable(np, 1);
+		fflp_errors_enable(np, 0);
+		nw64(0, H1POLY);
+		nw64(0, H2POLY);
+
+		err = tcam_flush_all(np);
+		if (err) {
+			niudbg(PROBE, "tcam_flush_all failed, err=%d\n",
+			       err);
+			goto out;
+		}
+		if (np->parent->plat_type == PLAT_TYPE_ATLAS) {
+			err = fflp_hash_clear(np);
+			if (err) {
+				niudbg(PROBE, "fflp_hash_clear failed, "
+				       "err=%d\n", err);
+				goto out;
+			}
+		}
+
+		vlan_tbl_clear(np);
+
+		niudbg(PROBE, "fflp_early_init: Success\n");
+		parent->flags |= PARENT_FLGS_CLS_HWINIT;
+	}
+out:
+	niu_unlock_parent(np, flags);
+	return err;
+}
+
+static int niu_set_flow_key(struct niu *np, unsigned long class_code, u64 key)
+{
+	if (class_code < CLASS_CODE_USER_PROG1 ||
+	    class_code > CLASS_CODE_SCTP_IPV6)
+		return -EINVAL;
+
+	nw64(key, FLOW_KEY(class_code - CLASS_CODE_USER_PROG1));
+	return 0;
+}
+
+static int niu_set_tcam_key(struct niu *np, unsigned long class_code, u64 key)
+{
+	if (class_code < CLASS_CODE_USER_PROG1 ||
+	    class_code > CLASS_CODE_SCTP_IPV6)
+		return -EINVAL;
+
+	nw64(key, TCAM_KEY(class_code - CLASS_CODE_USER_PROG1));
+	return 0;
+}
+
+static void niu_rx_skb_append(struct sk_buff *skb, struct page *page,
+			      __u32 offset, __u32 size)
+{
+	int i = skb_shinfo(skb)->nr_frags;
+	skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
+
+	frag->page = page;
+	frag->page_offset = offset;
+	frag->size = size;
+
+	skb->len += size;
+	skb->data_len += size;
+	skb->truesize += size;
+
+	skb_shinfo(skb)->nr_frags = i + 1;
+#if 0
+	{
+		unsigned char *data = page_address(page) + offset;
+		int q;
+
+		printk(KERN_ERR PFX "SKB append off(%u) size(%u) len(%u) dlen(%u) HDR [ ",
+		       offset, size, skb->len, skb->data_len);
+		for (q = 0; q < 14; q++)
+			printk("%02x", data[q]);
+		printk(" ]\n");
+	}
+#endif
+}
+
+static unsigned int niu_hash_rxaddr(struct rx_ring_info *rp, u64 a)
+{
+	a >>= PAGE_SHIFT;
+	a ^= (a >> ilog2(MAX_RBR_RING_SIZE));
+
+	return (a & (MAX_RBR_RING_SIZE - 1));
+}
+
+static struct page *niu_find_rxpage(struct rx_ring_info *rp, u64 addr,
+				    struct page ***link)
+{
+	unsigned int h = niu_hash_rxaddr(rp, addr);
+	struct page *p, **pp;
+
+	addr &= PAGE_MASK;
+	pp = &rp->rxhash[h];
+	for (; (p = *pp) != NULL; pp = (struct page **) &p->mapping) {
+		if (p->index == addr) {
+			*link = pp;
+			break;
+		}
+	}
+
+	return p;
+}
+
+static void niu_hash_page(struct rx_ring_info *rp, struct page *page, u64 base)
+{
+	unsigned int h = niu_hash_rxaddr(rp, base);
+
+	page->index = base;
+	page->mapping = (struct address_space *) rp->rxhash[h];
+	rp->rxhash[h] = page;
+}
+
+static int niu_rbr_add_page(struct niu *np, struct rx_ring_info *rp,
+			    gfp_t mask, int start_index)
+{
+	struct page *page;
+	u64 addr;
+	int i;
+
+	page = alloc_page(mask);
+	if (!page)
+		return -ENOMEM;
+
+	addr = np->ops->map_page(np->device, page, 0,
+				 PAGE_SIZE, DMA_FROM_DEVICE);
+
+	niu_hash_page(rp, page, addr);
+	if (rp->rbr_blocks_per_page > 1)
+		atomic_add(rp->rbr_blocks_per_page - 1,
+			   &compound_head(page)->_count);
+
+	for (i = 0; i < rp->rbr_blocks_per_page; i++) {
+		__le32 *rbr = &rp->rbr[start_index + i];
+
+		*rbr = cpu_to_le32(addr >> RBR_DESCR_ADDR_SHIFT);
+		addr += rp->rbr_block_size;
+	}
+
+	return 0;
+}
+
+static void niu_rbr_refill(struct niu *np, struct rx_ring_info *rp, gfp_t mask)
+{
+	int index = rp->rbr_index;
+
+	rp->rbr_pending++;
+	if ((rp->rbr_pending % rp->rbr_blocks_per_page) == 0) {
+		int err = niu_rbr_add_page(np, rp, mask, index);
+
+		if (unlikely(err)) {
+			rp->rbr_pending--;
+			return;
+		}
+
+		rp->rbr_index += rp->rbr_blocks_per_page;
+		BUG_ON(rp->rbr_index > rp->rbr_table_size);
+		if (rp->rbr_index == rp->rbr_table_size)
+			rp->rbr_index = 0;
+
+		if (rp->rbr_pending >= rp->rbr_kick_thresh) {
+			nw64(rp->rbr_pending, RBR_KICK(rp->rx_channel));
+			rp->rbr_pending = 0;
+		}
+	}
+}
+
+static int niu_rx_pkt_ignore(struct niu *np, struct rx_ring_info *rp)
+{
+	unsigned int index = rp->rcr_index;
+	int num_rcr = 0;
+
+	rp->rx_dropped++;
+	while (1) {
+		struct page *page, **link;
+		u64 addr, val;
+		u32 rcr_size;
+
+		num_rcr++;
+
+		val = le64_to_cpup(&rp->rcr[index]);
+		addr = (val & RCR_ENTRY_PKT_BUF_ADDR) <<
+			RCR_ENTRY_PKT_BUF_ADDR_SHIFT;
+		page = niu_find_rxpage(rp, addr, &link);
+		BUG_ON(!page);
+
+		rcr_size = rp->rbr_sizes[(val & RCR_ENTRY_PKTBUFSZ) >>
+					 RCR_ENTRY_PKTBUFSZ_SHIFT];
+		if ((page->index + PAGE_SIZE) - rcr_size == addr) {
+			*link = (struct page *) page->mapping;
+			np->ops->unmap_page(np->device, page->index,
+					    PAGE_SIZE, DMA_FROM_DEVICE);
+			page->index = 0;
+			page->mapping = NULL;
+			__free_page(page);
+			rp->rbr_refill_pending++;
+		}
+
+		index = NEXT_RCR(rp, index);
+		if (!(val & RCR_ENTRY_MULTI))
+			break;
+
+	}
+	rp->rcr_index = index;
+
+	return num_rcr;
+}
+
+static int niu_process_rx_pkt(struct niu *np, struct rx_ring_info *rp)
+{
+	unsigned int index = rp->rcr_index;
+	struct sk_buff *skb;
+	int len, num_rcr;
+
+	skb = netdev_alloc_skb(np->dev, RX_SKB_ALLOC_SIZE);
+	if (unlikely(!skb))
+		return niu_rx_pkt_ignore(np, rp);
+
+	num_rcr = 0;
+	while (1) {
+		struct page *page, **link;
+		u32 rcr_size, append_size;
+		u64 addr, val, off;
+
+		num_rcr++;
+
+		val = le64_to_cpup(&rp->rcr[index]);
+
+		len = (val & RCR_ENTRY_L2_LEN) >>
+			RCR_ENTRY_L2_LEN_SHIFT;
+		len -= ETH_FCS_LEN;
+
+		addr = (val & RCR_ENTRY_PKT_BUF_ADDR) <<
+			RCR_ENTRY_PKT_BUF_ADDR_SHIFT;
+		page = niu_find_rxpage(rp, addr, &link);
+		BUG_ON(!page);
+
+		rcr_size = rp->rbr_sizes[(val & RCR_ENTRY_PKTBUFSZ) >>
+					 RCR_ENTRY_PKTBUFSZ_SHIFT];
+
+		off = addr & ~PAGE_MASK;
+		append_size = rcr_size;
+		if (num_rcr == 1) {
+			int ptype;
+
+			off += 2;
+			append_size -= 2;
+
+			ptype = (val >> RCR_ENTRY_PKT_TYPE_SHIFT);
+			if ((ptype == RCR_PKT_TYPE_TCP ||
+			     ptype == RCR_PKT_TYPE_UDP) &&
+			    !(val & (RCR_ENTRY_NOPORT |
+				     RCR_ENTRY_ERROR)))
+				skb->ip_summed = CHECKSUM_UNNECESSARY;
+			else
+				skb->ip_summed = CHECKSUM_NONE;
+		}
+		if (!(val & RCR_ENTRY_MULTI))
+			append_size = len - skb->len;
+
+		niu_rx_skb_append(skb, page, off, append_size);
+		if ((page->index + rp->rbr_block_size) - rcr_size == addr) {
+			*link = (struct page *) page->mapping;
+			np->ops->unmap_page(np->device, page->index,
+					    PAGE_SIZE, DMA_FROM_DEVICE);
+			page->index = 0;
+			page->mapping = NULL;
+			rp->rbr_refill_pending++;
+		} else
+			get_page(page);
+
+		index = NEXT_RCR(rp, index);
+		if (!(val & RCR_ENTRY_MULTI))
+			break;
+
+	}
+	rp->rcr_index = index;
+
+	skb_reserve(skb, NET_IP_ALIGN);
+	__pskb_pull_tail(skb, min(len, NIU_RXPULL_MAX));
+
+	rp->rx_packets++;
+	rp->rx_bytes += skb->len;
+
+	skb->protocol = eth_type_trans(skb, np->dev);
+	netif_receive_skb(skb);
+
+	return num_rcr;
+}
+
+static int niu_rbr_fill(struct niu *np, struct rx_ring_info *rp, gfp_t mask)
+{
+	int blocks_per_page = rp->rbr_blocks_per_page;
+	int err, index = rp->rbr_index;
+
+	err = 0;
+	while (index < (rp->rbr_table_size - blocks_per_page)) {
+		err = niu_rbr_add_page(np, rp, mask, index);
+		if (err)
+			break;
+
+		index += blocks_per_page;
+	}
+
+	rp->rbr_index = index;
+	return err;
+}
+
+static void niu_rbr_free(struct niu *np, struct rx_ring_info *rp)
+{
+	int i;
+
+	for (i = 0; i < MAX_RBR_RING_SIZE; i++) {
+		struct page *page;
+
+		page = rp->rxhash[i];
+		while (page) {
+			struct page *next = (struct page *) page->mapping;
+			u64 base = page->index;
+
+			np->ops->unmap_page(np->device, base, PAGE_SIZE,
+					    DMA_FROM_DEVICE);
+			page->index = 0;
+			page->mapping = NULL;
+
+			__free_page(page);
+
+			page = next;
+		}
+	}
+
+	for (i = 0; i < rp->rbr_table_size; i++)
+		rp->rbr[i] = cpu_to_le32(0);
+	rp->rbr_index = 0;
+}
+
+static int release_tx_packet(struct niu *np, struct tx_ring_info *rp, int idx)
+{
+	struct tx_buff_info *tb = &rp->tx_buffs[idx];
+	struct sk_buff *skb = tb->skb;
+	struct tx_pkt_hdr *tp;
+	u64 tx_flags;
+	int i;
+
+	BUG_ON(!skb);
+
+	tp = (struct tx_pkt_hdr *) skb->data;
+	tx_flags = le64_to_cpup(&tp->flags);
+
+	rp->tx_packets++;
+	rp->tx_bytes += (((tx_flags & TXHDR_LEN) >> TXHDR_LEN_SHIFT) -
+			 ((tx_flags & TXHDR_PAD) / 2));
+
+	np->ops->unmap_single(np->device, tb->mapping,
+			      skb_headlen(skb), DMA_TO_DEVICE);
+
+	tb->skb = NULL;
+	idx = NEXT_TX(rp, idx);
+
+	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
+		tb = &rp->tx_buffs[idx];
+
+		BUG_ON(tb->skb != NULL);
+		np->ops->unmap_page(np->device, tb->mapping,
+				    skb_shinfo(skb)->frags[i].size,
+				    DMA_TO_DEVICE);
+		idx = NEXT_TX(rp, idx);
+	}
+
+	dev_kfree_skb(skb);
+
+	return idx;
+}
+
+#define NIU_TX_WAKEUP_THRESH(rp)		((rp)->pending / 4)
+
+static void niu_tx_work(struct niu *np, struct tx_ring_info *rp)
+{
+	u64 hdl, cs;
+	int head, cons;
+
+	cs = rp->tx_cs;
+	if (unlikely(!(cs & (TX_CS_MK | TX_CS_MMK))))
+		goto out;
+
+	hdl = rp->tx_hdl;
+	head = (hdl & TX_RING_HDL_HEAD) >> TX_RING_HDL_HEAD_SHIFT;
+	cons = rp->cons;
+
+	niudbg(TX_WORK, "%s: niu_tx_work() head[%d] cons[%d]\n",
+	       np->dev->name, head, cons);
+
+	while (cons != head)
+		cons = release_tx_packet(np, rp, cons);
+
+	rp->cons = cons;
+	smp_mb();
+
+out:
+	if (unlikely(netif_queue_stopped(np->dev) &&
+		     (niu_tx_avail(rp) > NIU_TX_WAKEUP_THRESH(rp)))) {
+		netif_tx_lock(np->dev);
+		if (netif_queue_stopped(np->dev) &&
+		    (niu_tx_avail(rp) > NIU_TX_WAKEUP_THRESH(rp)))
+			netif_wake_queue(np->dev);
+		netif_tx_unlock(np->dev);
+	}
+}
+
+static int niu_rx_work(struct niu *np, struct rx_ring_info *rp, int budget)
+{
+	int qlen, rcr_done = 0, work_done = 0;
+	struct rxdma_mailbox *mbox = rp->mbox;
+	u64 stat;
+
+#if 1
+	stat = nr64(RX_DMA_CTL_STAT(rp->rx_channel));
+	qlen = nr64(RCRSTAT_A(rp->rx_channel)) & RCRSTAT_A_QLEN;
+#else
+	stat = le64_to_cpup(&mbox->rx_dma_ctl_stat);
+	qlen = (le64_to_cpup(&mbox->rcrstat_a) & RCRSTAT_A_QLEN);
+#endif
+	mbox->rx_dma_ctl_stat = 0;
+	mbox->rcrstat_a = 0;
+
+	niudbg(RX_WORK, "%s: niu_rx_work(chan[%d]), stat[%lx] qlen=%d\n",
+	       np->dev->name, rp->rx_channel, stat, qlen);
+
+	rcr_done = work_done = 0;
+	qlen = min(qlen, budget);
+	while (work_done < qlen) {
+		rcr_done += niu_process_rx_pkt(np, rp);
+		work_done++;
+	}
+
+	if (rp->rbr_refill_pending >= rp->rbr_kick_thresh) {
+		unsigned int i;
+
+		for (i = 0; i < rp->rbr_refill_pending; i++)
+			niu_rbr_refill(np, rp, GFP_ATOMIC);
+		rp->rbr_refill_pending = 0;
+	}
+
+	stat = (RX_DMA_CTL_STAT_MEX |
+		((u64)work_done << RX_DMA_CTL_STAT_PKTREAD_SHIFT) |
+		((u64)rcr_done << RX_DMA_CTL_STAT_PTRREAD_SHIFT));
+
+	nw64(stat, RX_DMA_CTL_STAT(rp->rx_channel));
+
+	return work_done;
+}
+
+static int niu_poll_core(struct niu *np, struct niu_ldg *lp, int budget)
+{
+	u64 v0 = lp->v0;
+	u32 tx_vec = (v0 >> 32);
+	u32 rx_vec = (v0 & 0xffffffff);
+	int i, work_done = 0;
+
+	niudbg(POLL, "%s: niu_poll_core() v0[%016lx]\n",
+	       np->dev->name, v0);
+
+	for (i = 0; i < np->num_tx_rings; i++) {
+		struct tx_ring_info *rp = &np->tx_rings[i];
+		if (tx_vec & (1 << rp->tx_channel))
+			niu_tx_work(np, rp);
+		nw64(0, LD_IM0(LDN_TXDMA(rp->tx_channel)));
+	}
+
+	for (i = 0; i < np->num_rx_rings; i++) {
+		struct rx_ring_info *rp = &np->rx_rings[i];
+
+		if (rx_vec & (1 << rp->rx_channel)) {
+			int this_work_done;
+
+			this_work_done = niu_rx_work(np, rp,
+						     budget);
+
+			budget -= this_work_done;
+			work_done += this_work_done;
+		}
+		nw64(0, LD_IM0(LDN_RXDMA(rp->rx_channel)));
+	}
+
+	return work_done;
+}
+
+static int niu_poll(struct napi_struct *napi, int budget)
+{
+	struct niu_ldg *lp = container_of(napi, struct niu_ldg, napi);
+	struct niu *np = lp->np;
+	int work_done;
+
+	work_done = niu_poll_core(np, lp, budget);
+
+	if (work_done < budget) {
+		netif_rx_complete(np->dev, napi);
+		niu_ldg_rearm(np, lp, 1);
+	}
+	return work_done;
+}
+
+static void niu_log_rxchan_errors(struct niu *np, struct rx_ring_info *rp,
+				  u64 stat)
+{
+	printk(KERN_ERR PFX "%s: RX channel %u errors ( ",
+	       np->dev->name, rp->rx_channel);
+
+	if (stat & RX_DMA_CTL_STAT_RBR_TMOUT)
+		printk("RBR_TMOUT ");
+	if (stat & RX_DMA_CTL_STAT_RSP_CNT_ERR)
+		printk("RSP_CNT ");
+	if (stat & RX_DMA_CTL_STAT_BYTE_EN_BUS)
+		printk("BYTE_EN_BUS ");
+	if (stat & RX_DMA_CTL_STAT_RSP_DAT_ERR)
+		printk("RSP_DAT ");
+	if (stat & RX_DMA_CTL_STAT_RCR_ACK_ERR)
+		printk("RCR_ACK ");
+	if (stat & RX_DMA_CTL_STAT_RCR_SHA_PAR)
+		printk("RCR_SHA_PAR ");
+	if (stat & RX_DMA_CTL_STAT_RBR_PRE_PAR)
+		printk("RBR_PRE_PAR ");
+	if (stat & RX_DMA_CTL_STAT_CONFIG_ERR)
+		printk("CONFIG ");
+	if (stat & RX_DMA_CTL_STAT_RCRINCON)
+		printk("RCRINCON ");
+	if (stat & RX_DMA_CTL_STAT_RCRFULL)
+		printk("RCRFULL ");
+	if (stat & RX_DMA_CTL_STAT_RBRFULL)
+		printk("RBRFULL ");
+	if (stat & RX_DMA_CTL_STAT_RBRLOGPAGE)
+		printk("RBRLOGPAGE ");
+	if (stat & RX_DMA_CTL_STAT_CFIGLOGPAGE)
+		printk("CFIGLOGPAGE ");
+	if (stat & RX_DMA_CTL_STAT_DC_FIFO_ERR)
+		printk("DC_FIDO ");
+
+	printk(")\n");
+}
+
+static int niu_rx_error(struct niu *np, struct rx_ring_info *rp)
+{
+	u64 stat = nr64(RX_DMA_CTL_STAT(rp->rx_channel));
+	int err = 0;
+
+	printk(KERN_ERR PFX "%s: RX channel %u error, stat[%lx]\n",
+	       np->dev->name, rp->rx_channel, stat);
+
+	niu_log_rxchan_errors(np, rp, stat);
+
+	if (stat & (RX_DMA_CTL_STAT_CHAN_FATAL |
+		    RX_DMA_CTL_STAT_PORT_FATAL))
+		err = -EINVAL;
+
+	nw64(stat & RX_DMA_CTL_WRITE_CLEAR_ERRS,
+	     RX_DMA_CTL_STAT(rp->rx_channel));
+
+	return err;
+}
+
+static void niu_log_txchan_errors(struct niu *np, struct tx_ring_info *rp,
+				  u64 cs)
+{
+	printk(KERN_ERR PFX "%s: TX channel %u errors ( ",
+	       np->dev->name, rp->tx_channel);
+
+	if (cs & TX_CS_MBOX_ERR)
+		printk("MBOX ");
+	if (cs & TX_CS_PKT_SIZE_ERR)
+		printk("PKT_SIZE ");
+	if (cs & TX_CS_TX_RING_OFLOW)
+		printk("TX_RING_OFLOW ");
+	if (cs & TX_CS_PREF_BUF_PAR_ERR)
+		printk("PREF_BUF_PAR ");
+	if (cs