tehuti: Move the Tehuti driver

Move the Tehuti driver into drivers/net/ethernet/tehuti/ and
make the necessary Kconfig and Makefile changes.

CC: Alexander Indenbaum <baum@tehutinetworks.net>
CC: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

4 files changed, 3062 insertions, 0 deletions

diff --git a/drivers/net/ethernet/tehuti/Kconfig b/drivers/net/ethernet/tehuti/Kconfig
new file mode 100644
index 000000000000..914ad4059eae
--- /dev/null
+++ b/drivers/net/ethernet/tehuti/Kconfig
@@ -0,0 +1,26 @@
+#
+# Tehuti network device configuration
+#
+
+config NET_VENDOR_TEHUTI
+	bool "Tehuti devices"
+	depends on PCI
+	---help---
+	  If you have a network (Ethernet) card belonging to this class, say Y
+	  and read the Ethernet-HOWTO, available from
+	  <http://www.tldp.org/docs.html#howto>.
+
+	  Note that the answer to this question doesn't directly affect the
+	  kernel: saying N will just cause the configurator to skip all
+	  the questions about Tehuti cards. If you say Y, you will be asked for
+	  your specific card in the following questions.
+
+if NET_VENDOR_TEHUTI
+
+config TEHUTI
+	tristate "Tehuti Networks 10G Ethernet"
+	depends on PCI
+	---help---
+	  Tehuti Networks 10G Ethernet NIC
+
+endif # NET_VENDOR_TEHUTI
diff --git a/drivers/net/ethernet/tehuti/Makefile b/drivers/net/ethernet/tehuti/Makefile
new file mode 100644
index 000000000000..f995421ddbc8
--- /dev/null
+++ b/drivers/net/ethernet/tehuti/Makefile
@@ -0,0 +1,5 @@
+#
+# Makefile for the Tehuti network device drivers.
+#
+
+obj-$(CONFIG_TEHUTI) += tehuti.o
diff --git a/drivers/net/ethernet/tehuti/tehuti.c b/drivers/net/ethernet/tehuti/tehuti.c
new file mode 100644
index 000000000000..749bbf18dc6a
--- /dev/null
+++ b/drivers/net/ethernet/tehuti/tehuti.c
@@ -0,0 +1,2470 @@
+/*
+ * Tehuti Networks(R) Network Driver
+ * ethtool interface implementation
+ * Copyright (C) 2007 Tehuti Networks Ltd. All rights reserved
+ *
+ * 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.
+ */
+
+/*
+ * RX HW/SW interaction overview
+ * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+ * There are 2 types of RX communication channels between driver and NIC.
+ * 1) RX Free Fifo - RXF - holds descriptors of empty buffers to accept incoming
+ * traffic. This Fifo is filled by SW and is readen by HW. Each descriptor holds
+ * info about buffer's location, size and ID. An ID field is used to identify a
+ * buffer when it's returned with data via RXD Fifo (see below)
+ * 2) RX Data Fifo - RXD - holds descriptors of full buffers. This Fifo is
+ * filled by HW and is readen by SW. Each descriptor holds status and ID.
+ * HW pops descriptor from RXF Fifo, stores ID, fills buffer with incoming data,
+ * via dma moves it into host memory, builds new RXD descriptor with same ID,
+ * pushes it into RXD Fifo and raises interrupt to indicate new RX data.
+ *
+ * Current NIC configuration (registers + firmware) makes NIC use 2 RXF Fifos.
+ * One holds 1.5K packets and another - 26K packets. Depending on incoming
+ * packet size, HW desides on a RXF Fifo to pop buffer from. When packet is
+ * filled with data, HW builds new RXD descriptor for it and push it into single
+ * RXD Fifo.
+ *
+ * RX SW Data Structures
+ * ~~~~~~~~~~~~~~~~~~~~~
+ * skb db - used to keep track of all skbs owned by SW and their dma addresses.
+ * For RX case, ownership lasts from allocating new empty skb for RXF until
+ * accepting full skb from RXD and passing it to OS. Each RXF Fifo has its own
+ * skb db. Implemented as array with bitmask.
+ * fifo - keeps info about fifo's size and location, relevant HW registers,
+ * usage and skb db. Each RXD and RXF Fifo has its own fifo structure.
+ * Implemented as simple struct.
+ *
+ * RX SW Execution Flow
+ * ~~~~~~~~~~~~~~~~~~~~
+ * Upon initialization (ifconfig up) driver creates RX fifos and initializes
+ * relevant registers. At the end of init phase, driver enables interrupts.
+ * NIC sees that there is no RXF buffers and raises
+ * RD_INTR interrupt, isr fills skbs and Rx begins.
+ * Driver has two receive operation modes:
+ *    NAPI - interrupt-driven mixed with polling
+ *    interrupt-driven only
+ *
+ * Interrupt-driven only flow is following. When buffer is ready, HW raises
+ * interrupt and isr is called. isr collects all available packets
+ * (bdx_rx_receive), refills skbs (bdx_rx_alloc_skbs) and exit.
+
+ * Rx buffer allocation note
+ * ~~~~~~~~~~~~~~~~~~~~~~~~~
+ * Driver cares to feed such amount of RxF descriptors that respective amount of
+ * RxD descriptors can not fill entire RxD fifo. The main reason is lack of
+ * overflow check in Bordeaux for RxD fifo free/used size.
+ * FIXME: this is NOT fully implemented, more work should be done
+ *
+ */
+
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+
+#include "tehuti.h"
+
+static DEFINE_PCI_DEVICE_TABLE(bdx_pci_tbl) = {
+	{0x1FC9, 0x3009, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
+	{0x1FC9, 0x3010, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
+	{0x1FC9, 0x3014, PCI_ANY_ID, PCI_ANY_ID, 0, 0, 0},
+	{0}
+};
+
+MODULE_DEVICE_TABLE(pci, bdx_pci_tbl);
+
+/* Definitions needed by ISR or NAPI functions */
+static void bdx_rx_alloc_skbs(struct bdx_priv *priv, struct rxf_fifo *f);
+static void bdx_tx_cleanup(struct bdx_priv *priv);
+static int bdx_rx_receive(struct bdx_priv *priv, struct rxd_fifo *f, int budget);
+
+/* Definitions needed by FW loading */
+static void bdx_tx_push_desc_safe(struct bdx_priv *priv, void *data, int size);
+
+/* Definitions needed by hw_start */
+static int bdx_tx_init(struct bdx_priv *priv);
+static int bdx_rx_init(struct bdx_priv *priv);
+
+/* Definitions needed by bdx_close */
+static void bdx_rx_free(struct bdx_priv *priv);
+static void bdx_tx_free(struct bdx_priv *priv);
+
+/* Definitions needed by bdx_probe */
+static void bdx_set_ethtool_ops(struct net_device *netdev);
+
+/*************************************************************************
+ *    Print Info                                                         *
+ *************************************************************************/
+
+static void print_hw_id(struct pci_dev *pdev)
+{
+	struct pci_nic *nic = pci_get_drvdata(pdev);
+	u16 pci_link_status = 0;
+	u16 pci_ctrl = 0;
+
+	pci_read_config_word(pdev, PCI_LINK_STATUS_REG, &pci_link_status);
+	pci_read_config_word(pdev, PCI_DEV_CTRL_REG, &pci_ctrl);
+
+	pr_info("%s%s\n", BDX_NIC_NAME,
+		nic->port_num == 1 ? "" : ", 2-Port");
+	pr_info("srom 0x%x fpga %d build %u lane# %d max_pl 0x%x mrrs 0x%x\n",
+		readl(nic->regs + SROM_VER), readl(nic->regs + FPGA_VER) & 0xFFF,
+		readl(nic->regs + FPGA_SEED),
+		GET_LINK_STATUS_LANES(pci_link_status),
+		GET_DEV_CTRL_MAXPL(pci_ctrl), GET_DEV_CTRL_MRRS(pci_ctrl));
+}
+
+static void print_fw_id(struct pci_nic *nic)
+{
+	pr_info("fw 0x%x\n", readl(nic->regs + FW_VER));
+}
+
+static void print_eth_id(struct net_device *ndev)
+{
+	netdev_info(ndev, "%s, Port %c\n",
+		    BDX_NIC_NAME, (ndev->if_port == 0) ? 'A' : 'B');
+
+}
+
+/*************************************************************************
+ *    Code                                                               *
+ *************************************************************************/
+
+#define bdx_enable_interrupts(priv)	\
+	do { WRITE_REG(priv, regIMR, IR_RUN); } while (0)
+#define bdx_disable_interrupts(priv)	\
+	do { WRITE_REG(priv, regIMR, 0); } while (0)
+
+/* bdx_fifo_init
+ * create TX/RX descriptor fifo for host-NIC communication.
+ * 1K extra space is allocated at the end of the fifo to simplify
+ * processing of descriptors that wraps around fifo's end
+ * @priv - NIC private structure
+ * @f - fifo to initialize
+ * @fsz_type - fifo size type: 0-4KB, 1-8KB, 2-16KB, 3-32KB
+ * @reg_XXX - offsets of registers relative to base address
+ *
+ * Returns 0 on success, negative value on failure
+ *
+ */
+static int
+bdx_fifo_init(struct bdx_priv *priv, struct fifo *f, int fsz_type,
+	      u16 reg_CFG0, u16 reg_CFG1, u16 reg_RPTR, u16 reg_WPTR)
+{
+	u16 memsz = FIFO_SIZE * (1 << fsz_type);
+
+	memset(f, 0, sizeof(struct fifo));
+	/* pci_alloc_consistent gives us 4k-aligned memory */
+	f->va = pci_alloc_consistent(priv->pdev,
+				     memsz + FIFO_EXTRA_SPACE, &f->da);
+	if (!f->va) {
+		pr_err("pci_alloc_consistent failed\n");
+		RET(-ENOMEM);
+	}
+	f->reg_CFG0 = reg_CFG0;
+	f->reg_CFG1 = reg_CFG1;
+	f->reg_RPTR = reg_RPTR;
+	f->reg_WPTR = reg_WPTR;
+	f->rptr = 0;
+	f->wptr = 0;
+	f->memsz = memsz;
+	f->size_mask = memsz - 1;
+	WRITE_REG(priv, reg_CFG0, (u32) ((f->da & TX_RX_CFG0_BASE) | fsz_type));
+	WRITE_REG(priv, reg_CFG1, H32_64(f->da));
+
+	RET(0);
+}
+
+/* bdx_fifo_free - free all resources used by fifo
+ * @priv - NIC private structure
+ * @f - fifo to release
+ */
+static void bdx_fifo_free(struct bdx_priv *priv, struct fifo *f)
+{
+	ENTER;
+	if (f->va) {
+		pci_free_consistent(priv->pdev,
+				    f->memsz + FIFO_EXTRA_SPACE, f->va, f->da);
+		f->va = NULL;
+	}
+	RET();
+}
+
+/*
+ * bdx_link_changed - notifies OS about hw link state.
+ * @bdx_priv - hw adapter structure
+ */
+static void bdx_link_changed(struct bdx_priv *priv)
+{
+	u32 link = READ_REG(priv, regMAC_LNK_STAT) & MAC_LINK_STAT;
+
+	if (!link) {
+		if (netif_carrier_ok(priv->ndev)) {
+			netif_stop_queue(priv->ndev);
+			netif_carrier_off(priv->ndev);
+			netdev_err(priv->ndev, "Link Down\n");
+		}
+	} else {
+		if (!netif_carrier_ok(priv->ndev)) {
+			netif_wake_queue(priv->ndev);
+			netif_carrier_on(priv->ndev);
+			netdev_err(priv->ndev, "Link Up\n");
+		}
+	}
+}
+
+static void bdx_isr_extra(struct bdx_priv *priv, u32 isr)
+{
+	if (isr & IR_RX_FREE_0) {
+		bdx_rx_alloc_skbs(priv, &priv->rxf_fifo0);
+		DBG("RX_FREE_0\n");
+	}
+
+	if (isr & IR_LNKCHG0)
+		bdx_link_changed(priv);
+
+	if (isr & IR_PCIE_LINK)
+		netdev_err(priv->ndev, "PCI-E Link Fault\n");
+
+	if (isr & IR_PCIE_TOUT)
+		netdev_err(priv->ndev, "PCI-E Time Out\n");
+
+}
+
+/* bdx_isr - Interrupt Service Routine for Bordeaux NIC
+ * @irq - interrupt number
+ * @ndev - network device
+ * @regs - CPU registers
+ *
+ * Return IRQ_NONE if it was not our interrupt, IRQ_HANDLED - otherwise
+ *
+ * It reads ISR register to know interrupt reasons, and proceed them one by one.
+ * Reasons of interest are:
+ *    RX_DESC - new packet has arrived and RXD fifo holds its descriptor
+ *    RX_FREE - number of free Rx buffers in RXF fifo gets low
+ *    TX_FREE - packet was transmited and RXF fifo holds its descriptor
+ */
+
+static irqreturn_t bdx_isr_napi(int irq, void *dev)
+{
+	struct net_device *ndev = dev;
+	struct bdx_priv *priv = netdev_priv(ndev);
+	u32 isr;
+
+	ENTER;
+	isr = (READ_REG(priv, regISR) & IR_RUN);
+	if (unlikely(!isr)) {
+		bdx_enable_interrupts(priv);
+		return IRQ_NONE;	/* Not our interrupt */
+	}
+
+	if (isr & IR_EXTRA)
+		bdx_isr_extra(priv, isr);
+
+	if (isr & (IR_RX_DESC_0 | IR_TX_FREE_0)) {
+		if (likely(napi_schedule_prep(&priv->napi))) {
+			__napi_schedule(&priv->napi);
+			RET(IRQ_HANDLED);
+		} else {
+			/* NOTE: we get here if intr has slipped into window
+			 * between these lines in bdx_poll:
+			 *    bdx_enable_interrupts(priv);
+			 *    return 0;
+			 * currently intrs are disabled (since we read ISR),
+			 * and we have failed to register next poll.
+			 * so we read the regs to trigger chip
+			 * and allow further interupts. */
+			READ_REG(priv, regTXF_WPTR_0);
+			READ_REG(priv, regRXD_WPTR_0);
+		}
+	}
+
+	bdx_enable_interrupts(priv);
+	RET(IRQ_HANDLED);
+}
+
+static int bdx_poll(struct napi_struct *napi, int budget)
+{
+	struct bdx_priv *priv = container_of(napi, struct bdx_priv, napi);
+	int work_done;
+
+	ENTER;
+	bdx_tx_cleanup(priv);
+	work_done = bdx_rx_receive(priv, &priv->rxd_fifo0, budget);
+	if ((work_done < budget) ||
+	    (priv->napi_stop++ >= 30)) {
+		DBG("rx poll is done. backing to isr-driven\n");
+
+		/* from time to time we exit to let NAPI layer release
+		 * device lock and allow waiting tasks (eg rmmod) to advance) */
+		priv->napi_stop = 0;
+
+		napi_complete(napi);
+		bdx_enable_interrupts(priv);
+	}
+	return work_done;
+}
+
+/* bdx_fw_load - loads firmware to NIC
+ * @priv - NIC private structure
+ * Firmware is loaded via TXD fifo, so it must be initialized first.
+ * Firware must be loaded once per NIC not per PCI device provided by NIC (NIC
+ * can have few of them). So all drivers use semaphore register to choose one
+ * that will actually load FW to NIC.
+ */
+
+static int bdx_fw_load(struct bdx_priv *priv)
+{
+	const struct firmware *fw = NULL;
+	int master, i;
+	int rc;
+
+	ENTER;
+	master = READ_REG(priv, regINIT_SEMAPHORE);
+	if (!READ_REG(priv, regINIT_STATUS) && master) {
+		rc = request_firmware(&fw, "tehuti/bdx.bin", &priv->pdev->dev);
+		if (rc)
+			goto out;
+		bdx_tx_push_desc_safe(priv, (char *)fw->data, fw->size);
+		mdelay(100);
+	}
+	for (i = 0; i < 200; i++) {
+		if (READ_REG(priv, regINIT_STATUS)) {
+			rc = 0;
+			goto out;
+		}
+		mdelay(2);
+	}
+	rc = -EIO;
+out:
+	if (master)
+		WRITE_REG(priv, regINIT_SEMAPHORE, 1);
+	if (fw)
+		release_firmware(fw);
+
+	if (rc) {
+		netdev_err(priv->ndev, "firmware loading failed\n");
+		if (rc == -EIO)
+			DBG("VPC = 0x%x VIC = 0x%x INIT_STATUS = 0x%x i=%d\n",
+			    READ_REG(priv, regVPC),
+			    READ_REG(priv, regVIC),
+			    READ_REG(priv, regINIT_STATUS), i);
+		RET(rc);
+	} else {
+		DBG("%s: firmware loading success\n", priv->ndev->name);
+		RET(0);
+	}
+}
+
+static void bdx_restore_mac(struct net_device *ndev, struct bdx_priv *priv)
+{
+	u32 val;
+
+	ENTER;
+	DBG("mac0=%x mac1=%x mac2=%x\n",
+	    READ_REG(priv, regUNC_MAC0_A),
+	    READ_REG(priv, regUNC_MAC1_A), READ_REG(priv, regUNC_MAC2_A));
+
+	val = (ndev->dev_addr[0] << 8) | (ndev->dev_addr[1]);
+	WRITE_REG(priv, regUNC_MAC2_A, val);
+	val = (ndev->dev_addr[2] << 8) | (ndev->dev_addr[3]);
+	WRITE_REG(priv, regUNC_MAC1_A, val);
+	val = (ndev->dev_addr[4] << 8) | (ndev->dev_addr[5]);
+	WRITE_REG(priv, regUNC_MAC0_A, val);
+
+	DBG("mac0=%x mac1=%x mac2=%x\n",
+	    READ_REG(priv, regUNC_MAC0_A),
+	    READ_REG(priv, regUNC_MAC1_A), READ_REG(priv, regUNC_MAC2_A));
+	RET();
+}
+
+/* bdx_hw_start - inits registers and starts HW's Rx and Tx engines
+ * @priv - NIC private structure
+ */
+static int bdx_hw_start(struct bdx_priv *priv)
+{
+	int rc = -EIO;
+	struct net_device *ndev = priv->ndev;
+
+	ENTER;
+	bdx_link_changed(priv);
+
+	/* 10G overall max length (vlan, eth&ip header, ip payload, crc) */
+	WRITE_REG(priv, regFRM_LENGTH, 0X3FE0);
+	WRITE_REG(priv, regPAUSE_QUANT, 0x96);
+	WRITE_REG(priv, regRX_FIFO_SECTION, 0x800010);
+	WRITE_REG(priv, regTX_FIFO_SECTION, 0xE00010);
+	WRITE_REG(priv, regRX_FULLNESS, 0);
+	WRITE_REG(priv, regTX_FULLNESS, 0);
+	WRITE_REG(priv, regCTRLST,
+		  regCTRLST_BASE | regCTRLST_RX_ENA | regCTRLST_TX_ENA);
+
+	WRITE_REG(priv, regVGLB, 0);
+	WRITE_REG(priv, regMAX_FRAME_A,
+		  priv->rxf_fifo0.m.pktsz & MAX_FRAME_AB_VAL);
+
+	DBG("RDINTCM=%08x\n", priv->rdintcm);	/*NOTE: test script uses this */
+	WRITE_REG(priv, regRDINTCM0, priv->rdintcm);
+	WRITE_REG(priv, regRDINTCM2, 0);	/*cpu_to_le32(rcm.val)); */
+
+	DBG("TDINTCM=%08x\n", priv->tdintcm);	/*NOTE: test script uses this */
+	WRITE_REG(priv, regTDINTCM0, priv->tdintcm);	/* old val = 0x300064 */
+
+	/* Enable timer interrupt once in 2 secs. */
+	/*WRITE_REG(priv, regGTMR0, ((GTMR_SEC * 2) & GTMR_DATA)); */
+	bdx_restore_mac(priv->ndev, priv);
+
+	WRITE_REG(priv, regGMAC_RXF_A, GMAC_RX_FILTER_OSEN |
+		  GMAC_RX_FILTER_AM | GMAC_RX_FILTER_AB);
+
+#define BDX_IRQ_TYPE	((priv->nic->irq_type == IRQ_MSI) ? 0 : IRQF_SHARED)
+
+	rc = request_irq(priv->pdev->irq, bdx_isr_napi, BDX_IRQ_TYPE,
+			 ndev->name, ndev);
+	if (rc)
+		goto err_irq;
+	bdx_enable_interrupts(priv);
+
+	RET(0);
+
+err_irq:
+	RET(rc);
+}
+
+static void bdx_hw_stop(struct bdx_priv *priv)
+{
+	ENTER;
+	bdx_disable_interrupts(priv);
+	free_irq(priv->pdev->irq, priv->ndev);
+
+	netif_carrier_off(priv->ndev);
+	netif_stop_queue(priv->ndev);
+
+	RET();
+}
+
+static int bdx_hw_reset_direct(void __iomem *regs)
+{
+	u32 val, i;
+	ENTER;
+
+	/* reset sequences: read, write 1, read, write 0 */
+	val = readl(regs + regCLKPLL);
+	writel((val | CLKPLL_SFTRST) + 0x8, regs + regCLKPLL);
+	udelay(50);
+	val = readl(regs + regCLKPLL);
+	writel(val & ~CLKPLL_SFTRST, regs + regCLKPLL);
+
+	/* check that the PLLs are locked and reset ended */
+	for (i = 0; i < 70; i++, mdelay(10))
+		if ((readl(regs + regCLKPLL) & CLKPLL_LKD) == CLKPLL_LKD) {
+			/* do any PCI-E read transaction */
+			readl(regs + regRXD_CFG0_0);
+			return 0;
+		}
+	pr_err("HW reset failed\n");
+	return 1;		/* failure */
+}
+
+static int bdx_hw_reset(struct bdx_priv *priv)
+{
+	u32 val, i;
+	ENTER;
+
+	if (priv->port == 0) {
+		/* reset sequences: read, write 1, read, write 0 */
+		val = READ_REG(priv, regCLKPLL);
+		WRITE_REG(priv, regCLKPLL, (val | CLKPLL_SFTRST) + 0x8);
+		udelay(50);
+		val = READ_REG(priv, regCLKPLL);
+		WRITE_REG(priv, regCLKPLL, val & ~CLKPLL_SFTRST);
+	}
+	/* check that the PLLs are locked and reset ended */
+	for (i = 0; i < 70; i++, mdelay(10))
+		if ((READ_REG(priv, regCLKPLL) & CLKPLL_LKD) == CLKPLL_LKD) {
+			/* do any PCI-E read transaction */
+			READ_REG(priv, regRXD_CFG0_0);
+			return 0;
+		}
+	pr_err("HW reset failed\n");
+	return 1;		/* failure */
+}
+
+static int bdx_sw_reset(struct bdx_priv *priv)
+{
+	int i;
+
+	ENTER;
+	/* 1. load MAC (obsolete) */
+	/* 2. disable Rx (and Tx) */
+	WRITE_REG(priv, regGMAC_RXF_A, 0);
+	mdelay(100);
+	/* 3. disable port */
+	WRITE_REG(priv, regDIS_PORT, 1);
+	/* 4. disable queue */
+	WRITE_REG(priv, regDIS_QU, 1);
+	/* 5. wait until hw is disabled */
+	for (i = 0; i < 50; i++) {
+		if (READ_REG(priv, regRST_PORT) & 1)
+			break;
+		mdelay(10);
+	}
+	if (i == 50)
+		netdev_err(priv->ndev, "SW reset timeout. continuing anyway\n");
+
+	/* 6. disable intrs */
+	WRITE_REG(priv, regRDINTCM0, 0);
+	WRITE_REG(priv, regTDINTCM0, 0);
+	WRITE_REG(priv, regIMR, 0);
+	READ_REG(priv, regISR);
+
+	/* 7. reset queue */
+	WRITE_REG(priv, regRST_QU, 1);
+	/* 8. reset port */
+	WRITE_REG(priv, regRST_PORT, 1);
+	/* 9. zero all read and write pointers */
+	for (i = regTXD_WPTR_0; i <= regTXF_RPTR_3; i += 0x10)
+		DBG("%x = %x\n", i, READ_REG(priv, i) & TXF_WPTR_WR_PTR);
+	for (i = regTXD_WPTR_0; i <= regTXF_RPTR_3; i += 0x10)
+		WRITE_REG(priv, i, 0);
+	/* 10. unseet port disable */
+	WRITE_REG(priv, regDIS_PORT, 0);
+	/* 11. unset queue disable */
+	WRITE_REG(priv, regDIS_QU, 0);
+	/* 12. unset queue reset */
+	WRITE_REG(priv, regRST_QU, 0);
+	/* 13. unset port reset */
+	WRITE_REG(priv, regRST_PORT, 0);
+	/* 14. enable Rx */
+	/* skiped. will be done later */
+	/* 15. save MAC (obsolete) */
+	for (i = regTXD_WPTR_0; i <= regTXF_RPTR_3; i += 0x10)
+		DBG("%x = %x\n", i, READ_REG(priv, i) & TXF_WPTR_WR_PTR);
+
+	RET(0);
+}
+
+/* bdx_reset - performs right type of reset depending on hw type */
+static int bdx_reset(struct bdx_priv *priv)
+{
+	ENTER;
+	RET((priv->pdev->device == 0x3009)
+	    ? bdx_hw_reset(priv)
+	    : bdx_sw_reset(priv));
+}
+
+/**
+ * bdx_close - Disables a network interface
+ * @netdev: network interface device structure
+ *
+ * Returns 0, this is not allowed to fail
+ *
+ * The close entry point is called when an interface is de-activated
+ * by the OS.  The hardware is still under the drivers control, but
+ * needs to be disabled.  A global MAC reset is issued to stop the
+ * hardware, and all transmit and receive resources are freed.
+ **/
+static int bdx_close(struct net_device *ndev)
+{
+	struct bdx_priv *priv = NULL;
+
+	ENTER;
+	priv = netdev_priv(ndev);
+
+	napi_disable(&priv->napi);
+
+	bdx_reset(priv);
+	bdx_hw_stop(priv);
+	bdx_rx_free(priv);
+	bdx_tx_free(priv);
+	RET(0);
+}
+
+/**
+ * bdx_open - Called when a network interface is made active
+ * @netdev: network interface device structure
+ *
+ * Returns 0 on success, negative value on failure
+ *
+ * The open entry point is called when a network interface is made
+ * active by the system (IFF_UP).  At this point all resources needed
+ * for transmit and receive operations are allocated, the interrupt
+ * handler is registered with the OS, the watchdog timer is started,
+ * and the stack is notified that the interface is ready.
+ **/
+static int bdx_open(struct net_device *ndev)
+{
+	struct bdx_priv *priv;
+	int rc;
+
+	ENTER;
+	priv = netdev_priv(ndev);
+	bdx_reset(priv);
+	if (netif_running(ndev))
+		netif_stop_queue(priv->ndev);
+
+	if ((rc = bdx_tx_init(priv)) ||
+	    (rc = bdx_rx_init(priv)) ||
+	    (rc = bdx_fw_load(priv)))
+		goto err;
+
+	bdx_rx_alloc_skbs(priv, &priv->rxf_fifo0);
+
+	rc = bdx_hw_start(priv);
+	if (rc)
+		goto err;
+
+	napi_enable(&priv->napi);
+
+	print_fw_id(priv->nic);
+
+	RET(0);
+
+err:
+	bdx_close(ndev);
+	RET(rc);
+}
+
+static int bdx_range_check(struct bdx_priv *priv, u32 offset)
+{
+	return (offset > (u32) (BDX_REGS_SIZE / priv->nic->port_num)) ?
+		-EINVAL : 0;
+}
+
+static int bdx_ioctl_priv(struct net_device *ndev, struct ifreq *ifr, int cmd)
+{
+	struct bdx_priv *priv = netdev_priv(ndev);
+	u32 data[3];
+	int error;
+
+	ENTER;
+
+	DBG("jiffies=%ld cmd=%d\n", jiffies, cmd);
+	if (cmd != SIOCDEVPRIVATE) {
+		error = copy_from_user(data, ifr->ifr_data, sizeof(data));
+		if (error) {
+			pr_err("can't copy from user\n");
+			RET(-EFAULT);
+		}
+		DBG("%d 0x%x 0x%x\n", data[0], data[1], data[2]);
+	}
+
+	if (!capable(CAP_SYS_RAWIO))
+		return -EPERM;
+
+	switch (data[0]) {
+
+	case BDX_OP_READ:
+		error = bdx_range_check(priv, data[1]);
+		if (error < 0)
+			return error;
+		data[2] = READ_REG(priv, data[1]);
+		DBG("read_reg(0x%x)=0x%x (dec %d)\n", data[1], data[2],
+		    data[2]);
+		error = copy_to_user(ifr->ifr_data, data, sizeof(data));
+		if (error)
+			RET(-EFAULT);
+		break;
+
+	case BDX_OP_WRITE:
+		error = bdx_range_check(priv, data[1]);
+		if (error < 0)
+			return error;
+		WRITE_REG(priv, data[1], data[2]);
+		DBG("write_reg(0x%x, 0x%x)\n", data[1], data[2]);
+		break;
+
+	default:
+		RET(-EOPNOTSUPP);
+	}
+	return 0;
+}
+
+static int bdx_ioctl(struct net_device *ndev, struct ifreq *ifr, int cmd)
+{
+	ENTER;
+	if (cmd >= SIOCDEVPRIVATE && cmd <= (SIOCDEVPRIVATE + 15))
+		RET(bdx_ioctl_priv(ndev, ifr, cmd));
+	else
+		RET(-EOPNOTSUPP);
+}
+
+/*
+ * __bdx_vlan_rx_vid - private helper for adding/killing VLAN vid
+ *                     by passing VLAN filter table to hardware
+ * @ndev network device
+ * @vid  VLAN vid
+ * @op   add or kill operation
+ */
+static void __bdx_vlan_rx_vid(struct net_device *ndev, uint16_t vid, int enable)
+{
+	struct bdx_priv *priv = netdev_priv(ndev);
+	u32 reg, bit, val;
+
+	ENTER;
+	DBG2("vid=%d value=%d\n", (int)vid, enable);
+	if (unlikely(vid >= 4096)) {
+		pr_err("invalid VID: %u (> 4096)\n", vid);
+		RET();
+	}
+	reg = regVLAN_0 + (vid / 32) * 4;
+	bit = 1 << vid % 32;
+	val = READ_REG(priv, reg);
+	DBG2("reg=%x, val=%x, bit=%d\n", reg, val, bit);
+	if (enable)
+		val |= bit;
+	else
+		val &= ~bit;
+	DBG2("new val %x\n", val);
+	WRITE_REG(priv, reg, val);
+	RET();
+}
+
+/*
+ * bdx_vlan_rx_add_vid - kernel hook for adding VLAN vid to hw filtering table
+ * @ndev network device
+ * @vid  VLAN vid to add
+ */
+static void bdx_vlan_rx_add_vid(struct net_device *ndev, uint16_t vid)
+{
+	__bdx_vlan_rx_vid(ndev, vid, 1);
+}
+
+/*
+ * bdx_vlan_rx_kill_vid - kernel hook for killing VLAN vid in hw filtering table
+ * @ndev network device
+ * @vid  VLAN vid to kill
+ */
+static void bdx_vlan_rx_kill_vid(struct net_device *ndev, unsigned short vid)
+{
+	__bdx_vlan_rx_vid(ndev, vid, 0);
+}
+
+/**
+ * bdx_change_mtu - Change the Maximum Transfer Unit
+ * @netdev: network interface device structure
+ * @new_mtu: new value for maximum frame size
+ *
+ * Returns 0 on success, negative on failure
+ */
+static int bdx_change_mtu(struct net_device *ndev, int new_mtu)
+{
+	ENTER;
+
+	if (new_mtu == ndev->mtu)
+		RET(0);
+
+	/* enforce minimum frame size */
+	if (new_mtu < ETH_ZLEN) {
+		netdev_err(ndev, "mtu %d is less then minimal %d\n",
+			   new_mtu, ETH_ZLEN);
+		RET(-EINVAL);
+	}
+
+	ndev->mtu = new_mtu;
+	if (netif_running(ndev)) {
+		bdx_close(ndev);
+		bdx_open(ndev);
+	}
+	RET(0);
+}
+
+static void bdx_setmulti(struct net_device *ndev)
+{
+	struct bdx_priv *priv = netdev_priv(ndev);
+
+	u32 rxf_val =
+	    GMAC_RX_FILTER_AM | GMAC_RX_FILTER_AB | GMAC_RX_FILTER_OSEN;
+	int i;
+
+	ENTER;
+	/* IMF - imperfect (hash) rx multicat filter */
+	/* PMF - perfect rx multicat filter */
+
+	/* FIXME: RXE(OFF) */
+	if (ndev->flags & IFF_PROMISC) {
+		rxf_val |= GMAC_RX_FILTER_PRM;
+	} else if (ndev->flags & IFF_ALLMULTI) {
+		/* set IMF to accept all multicast frmaes */
+		for (i = 0; i < MAC_MCST_HASH_NUM; i++)
+			WRITE_REG(priv, regRX_MCST_HASH0 + i * 4, ~0);
+	} else if (!netdev_mc_empty(ndev)) {
+		u8 hash;
+		struct netdev_hw_addr *ha;
+		u32 reg, val;
+
+		/* set IMF to deny all multicast frames */
+		for (i = 0; i < MAC_MCST_HASH_NUM; i++)
+			WRITE_REG(priv, regRX_MCST_HASH0 + i * 4, 0);
+		/* set PMF to deny all multicast frames */
+		for (i = 0; i < MAC_MCST_NUM; i++) {
+			WRITE_REG(priv, regRX_MAC_MCST0 + i * 8, 0);
+			WRITE_REG(priv, regRX_MAC_MCST1 + i * 8, 0);
+		}
+
+		/* use PMF to accept first MAC_MCST_NUM (15) addresses */
+		/* TBD: sort addresses and write them in ascending order
+		 * into RX_MAC_MCST regs. we skip this phase now and accept ALL
+		 * multicast frames throu IMF */
+		/* accept the rest of addresses throu IMF */
+		netdev_for_each_mc_addr(ha, ndev) {
+			hash = 0;
+			for (i = 0; i < ETH_ALEN; i++)
+				hash ^= ha->addr[i];
+			reg = regRX_MCST_HASH0 + ((hash >> 5) << 2);
+			val = READ_REG(priv, reg);
+			val |= (1 << (hash % 32));
+			WRITE_REG(priv, reg, val);
+		}
+
+	} else {
+		DBG("only own mac %d\n", netdev_mc_count(ndev));
+		rxf_val |= GMAC_RX_FILTER_AB;
+	}
+	WRITE_REG(priv, regGMAC_RXF_A, rxf_val);
+	/* enable RX */
+	/* FIXME: RXE(ON) */
+	RET();
+}
+
+static int bdx_set_mac(struct net_device *ndev, void *p)
+{
+	struct bdx_priv *priv = netdev_priv(ndev);
+	struct sockaddr *addr = p;
+
+	ENTER;
+	/*
+	   if (netif_running(dev))
+	   return -EBUSY
+	 */
+	memcpy(ndev->dev_addr, addr->sa_data, ndev->addr_len);
+	bdx_restore_mac(ndev, priv);
+	RET(0);
+}
+
+static int bdx_read_mac(struct bdx_priv *priv)
+{
+	u16 macAddress[3], i;
+	ENTER;
+
+	macAddress[2] = READ_REG(priv, regUNC_MAC0_A);
+	macAddress[2] = READ_REG(priv, regUNC_MAC0_A);
+	macAddress[1] = READ_REG(priv, regUNC_MAC1_A);
+	macAddress[1] = READ_REG(priv, regUNC_MAC1_A);
+	macAddress[0] = READ_REG(priv, regUNC_MAC2_A);
+	macAddress[0] = READ_REG(priv, regUNC_MAC2_A);
+	for (i = 0; i < 3; i++) {
+		priv->ndev->dev_addr[i * 2 + 1] = macAddress[i];
+		priv->ndev->dev_addr[i * 2] = macAddress[i] >> 8;
+	}
+	RET(0);
+}
+
+static u64 bdx_read_l2stat(struct bdx_priv *priv, int reg)
+{
+	u64 val;
+
+	val = READ_REG(priv, reg);
+	val |= ((u64) READ_REG(priv, reg + 8)) << 32;
+	return val;
+}
+
+/*Do the statistics-update work*/
+static void bdx_update_stats(struct bdx_priv *priv)
+{
+	struct bdx_stats *stats = &priv->hw_stats;
+	u64 *stats_vector = (u64 *) stats;
+	int i;
+	int addr;
+
+	/*Fill HW structure */
+	addr = 0x7200;
+	/*First 12 statistics - 0x7200 - 0x72B0 */
+	for (i = 0; i < 12; i++) {
+		stats_vector[i] = bdx_read_l2stat(priv, addr);
+		addr += 0x10;
+	}
+	BDX_ASSERT(addr != 0x72C0);
+	/* 0x72C0-0x72E0 RSRV */
+	addr = 0x72F0;
+	for (; i < 16; i++) {
+		stats_vector[i] = bdx_read_l2stat(priv, addr);
+		addr += 0x10;
+	}
+	BDX_ASSERT(addr != 0x7330);
+	/* 0x7330-0x7360 RSRV */
+	addr = 0x7370;
+	for (; i < 19; i++) {
+		stats_vector[i] = bdx_read_l2stat(priv, addr);
+		addr += 0x10;
+	}
+	BDX_ASSERT(addr != 0x73A0);
+	/* 0x73A0-0x73B0 RSRV */
+	addr = 0x73C0;
+	for (; i < 23; i++) {
+		stats_vector[i] = bdx_read_l2stat(priv, addr);
+		addr += 0x10;
+	}
+	BDX_ASSERT(addr != 0x7400);
+	BDX_ASSERT((sizeof(struct bdx_stats) / sizeof(u64)) != i);
+}
+
+static void print_rxdd(struct rxd_desc *rxdd, u32 rxd_val1, u16 len,
+		       u16 rxd_vlan);
+static void print_rxfd(struct rxf_desc *rxfd);
+
+/*************************************************************************
+ *     Rx DB                                                             *
+ *************************************************************************/
+
+static void bdx_rxdb_destroy(struct rxdb *db)
+{
+	vfree(db);
+}
+
+static struct rxdb *bdx_rxdb_create(int nelem)
+{
+	struct rxdb *db;
+	int i;
+
+	db = vmalloc(sizeof(struct rxdb)
+		     + (nelem * sizeof(int))
+		     + (nelem * sizeof(struct rx_map)));
+	if (likely(db != NULL)) {
+		db->stack = (int *)(db + 1);
+		db->elems = (void *)(db->stack + nelem);
+		db->nelem = nelem;
+		db->top = nelem;
+		for (i = 0; i < nelem; i++)
+			db->stack[i] = nelem - i - 1;	/* to make first allocs
+							   close to db struct*/
+	}
+
+	return db;
+}
+
+static inline int bdx_rxdb_alloc_elem(struct rxdb *db)
+{
+	BDX_ASSERT(db->top <= 0);
+	return db->stack[--(db->top)];
+}
+
+static inline void *bdx_rxdb_addr_elem(struct rxdb *db, int n)
+{
+	BDX_ASSERT((n < 0) || (n >= db->nelem));
+	return db->elems + n;
+}
+
+static inline int bdx_rxdb_available(struct rxdb *db)
+{
+	return db->top;
+}
+
+static inline void bdx_rxdb_free_elem(struct rxdb *db, int n)
+{
+	BDX_ASSERT((n >= db->nelem) || (n < 0));
+	db->stack[(db->top)++] = n;
+}
+
+/*************************************************************************
+ *     Rx Init                                                           *
+ *************************************************************************/
+
+/* bdx_rx_init - initialize RX all related HW and SW resources
+ * @priv - NIC private structure
+ *
+ * Returns 0 on success, negative value on failure
+ *
+ * It creates rxf and rxd fifos, update relevant HW registers, preallocate
+ * skb for rx. It assumes that Rx is desabled in HW
+ * funcs are grouped for better cache usage
+ *
+ * RxD fifo is smaller than RxF fifo by design. Upon high load, RxD will be
+ * filled and packets will be dropped by nic without getting into host or
+ * cousing interrupt. Anyway, in that condition, host has no chance to process
+ * all packets, but dropping in nic is cheaper, since it takes 0 cpu cycles
+ */
+
+/* TBD: ensure proper packet size */
+
+static int bdx_rx_init(struct bdx_priv *priv)
+{
+	ENTER;
+
+	if (bdx_fifo_init(priv, &priv->rxd_fifo0.m, priv->rxd_size,
+			  regRXD_CFG0_0, regRXD_CFG1_0,
+			  regRXD_RPTR_0, regRXD_WPTR_0))
+		goto err_mem;
+	if (bdx_fifo_init(priv, &priv->rxf_fifo0.m, priv->rxf_size,
+			  regRXF_CFG0_0, regRXF_CFG1_0,
+			  regRXF_RPTR_0, regRXF_WPTR_0))
+		goto err_mem;
+	priv->rxdb = bdx_rxdb_create(priv->rxf_fifo0.m.memsz /
+				     sizeof(struct rxf_desc));
+	if (!priv->rxdb)
+		goto err_mem;
+
+	priv->rxf_fifo0.m.pktsz = priv->ndev->mtu + VLAN_ETH_HLEN;
+	return 0;
+
+err_mem:
+	netdev_err(priv->ndev, "Rx init failed\n");
+	return -ENOMEM;
+}
+
+/* bdx_rx_free_skbs - frees and unmaps all skbs allocated for the fifo
+ * @priv - NIC private structure
+ * @f - RXF fifo
+ */
+static void bdx_rx_free_skbs(struct bdx_priv *priv, struct rxf_fifo *f)
+{
+	struct rx_map *dm;
+	struct rxdb *db = priv->rxdb;
+	u16 i;
+
+	ENTER;
+	DBG("total=%d free=%d busy=%d\n", db->nelem, bdx_rxdb_available(db),
+	    db->nelem - bdx_rxdb_available(db));
+	while (bdx_rxdb_available(db) > 0) {
+		i = bdx_rxdb_alloc_elem(db);
+		dm = bdx_rxdb_addr_elem(db, i);
+		dm->dma = 0;
+	}
+	for (i = 0; i < db->nelem; i++) {
+		dm = bdx_rxdb_addr_elem(db, i);
+		if (dm->dma) {
+			pci_unmap_single(priv->pdev,
+					 dm->dma, f->m.pktsz,
+					 PCI_DMA_FROMDEVICE);
+			dev_kfree_skb(dm->skb);
+		}
+	}
+}
+
+/* bdx_rx_free - release all Rx resources
+ * @priv - NIC private structure
+ * It assumes that Rx is desabled in HW
+ */
+static void bdx_rx_free(struct bdx_priv *priv)
+{
+	ENTER;
+	if (priv->rxdb) {
+		bdx_rx_free_skbs(priv, &priv->rxf_fifo0);
+		bdx_rxdb_destroy(priv->rxdb);
+		priv->rxdb = NULL;
+	}
+	bdx_fifo_free(priv, &priv->rxf_fifo0.m);
+	bdx_fifo_free(priv, &priv->rxd_fifo0.m);
+
+	RET();
+}
+
+/*************************************************************************