Adding upstream version 6.1.76.upstream/6.1.76upstream

Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>

4 files changed, 3049 insertions, 0 deletions

diff --git a/drivers/net/ethernet/tehuti/Kconfig b/drivers/net/ethernet/tehuti/Kconfig
new file mode 100644
index 000000000..873563376
--- /dev/null
+++ b/drivers/net/ethernet/tehuti/Kconfig
@@ -0,0 +1,26 @@
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# Tehuti network device configuration
+#
+
+config NET_VENDOR_TEHUTI
+	bool "Tehuti devices"
+	default y
+	depends on PCI
+	help
+	  If you have a network (Ethernet) card belonging to this class, say Y.
+
+	  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 000000000..13a0ddd62
--- /dev/null
+++ b/drivers/net/ethernet/tehuti/Makefile
@@ -0,0 +1,6 @@
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# 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 000000000..ca409515e
--- /dev/null
+++ b/drivers/net/ethernet/tehuti/tehuti.c
@@ -0,0 +1,2457 @@
+// SPDX-License-Identifier: GPL-2.0-or-later
+/*
+ * Tehuti Networks(R) Network Driver
+ * ethtool interface implementation
+ * Copyright (C) 2007 Tehuti Networks Ltd. All rights reserved
+ */
+
+/*
+ * 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 const struct pci_device_id bdx_pci_tbl[] = {
+	{ PCI_VDEVICE(TEHUTI, 0x3009), },
+	{ PCI_VDEVICE(TEHUTI, 0x3010), },
+	{ PCI_VDEVICE(TEHUTI, 0x3014), },
+	{ 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.
+ * @priv: NIC private structure
+ * @f: fifo to initialize
+ * @fsz_type: fifo size type: 0-4KB, 1-8KB, 2-16KB, 3-32KB
+ * @reg_CFG0: offsets of registers relative to base address
+ * @reg_CFG1: offsets of registers relative to base address
+ * @reg_RPTR: offsets of registers relative to base address
+ * @reg_WPTR: offsets of registers relative to base address
+ *
+ * 1K extra space is allocated at the end of the fifo to simplify
+ * processing of descriptors that wraps around fifo's end
+ *
+ * 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));
+	/* dma_alloc_coherent gives us 4k-aligned memory */
+	f->va = dma_alloc_coherent(&priv->pdev->dev, memsz + FIFO_EXTRA_SPACE,
+				   &f->da, GFP_ATOMIC);
+	if (!f->va) {
+		pr_err("dma_alloc_coherent 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) {
+		dma_free_coherent(&priv->pdev->dev,
+				  f->memsz + FIFO_EXTRA_SPACE, f->va, f->da);
+		f->va = NULL;
+	}
+	RET();
+}
+
+/**
+ * bdx_link_changed - notifies OS about hw link state.
+ * @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_napi - Interrupt Service Routine for Bordeaux NIC
+ * @irq: interrupt number
+ * @dev: network device
+ *
+ * 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_done(napi, work_done);
+		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);
+
+	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
+ * @ndev: 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
+ * @ndev: 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_siocdevprivate(struct net_device *ndev, struct ifreq *ifr,
+			      void __user *udata, 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, udata, 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]);
+	} else {
+		return -EOPNOTSUPP;
+	}
+
+	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(udata, 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;
+}
+
+/**
+ * __bdx_vlan_rx_vid - private helper for adding/killing VLAN vid
+ * @ndev: network device
+ * @vid:  VLAN vid
+ * @enable: enable or disable vlan
+ *
+ * Passes VLAN filter table to hardware
+ */
+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
+ * @proto: unused
+ * @vid:  VLAN vid to add
+ */
+static int bdx_vlan_rx_add_vid(struct net_device *ndev, __be16 proto, u16 vid)
+{
+	__bdx_vlan_rx_vid(ndev, vid, 1);
+	return 0;
+}
+
+/**
+ * bdx_vlan_rx_kill_vid - kernel hook for killing VLAN vid in hw filtering table
+ * @ndev: network device
+ * @proto: unused
+ * @vid:  VLAN vid to kill
+ */
+static int bdx_vlan_rx_kill_vid(struct net_device *ndev, __be16 proto, u16 vid)
+{
+	__bdx_vlan_rx_vid(ndev, vid, 0);
+	return 0;
+}
+
+/**
+ * bdx_change_mtu - Change the Maximum Transfer Unit
+ * @ndev: 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;
+
+	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
+	 */
+	eth_hw_addr_set(ndev, addr->sa_data);
+	bdx_restore_mac(ndev, priv);
+	RET(0);
+}
+
+static int bdx_read_mac(struct bdx_priv *priv)
+{
+	u16 macAddress[3], i;
+	u8 addr[ETH_ALEN];
+	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++) {
+		addr[i * 2 + 1] = macAddress[i];
+		addr[i * 2] = macAddress[i] >> 8;
+	}
+	eth_hw_addr_set(priv->ndev, addr);
+	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) {
+			dma_unmap_single(&priv->pdev->dev, dm->dma,
+					 f->m.pktsz, DMA_FROM_DEVICE);
+			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();
+}
+
+/*************************************************************************
+ *     Rx Engine                                                         *
+ *************************************************************************/
+
+/**
+ * bdx_rx_alloc_skbs - fill rxf fifo with new skbs
+ * @priv: nic's private structure
+ * @f: RXF fifo that needs skbs
+ *
+ * It allocates skbs, build rxf descs and push it (rxf descr) into rxf fifo.
+ * skb's virtual and physical addresses are stored in skb db.
+ * To calculate free space, func uses cached values of RPTR and WPTR
+ * When needed, it also updates RPTR and WPTR.
+ */
+
+/* TBD: do not update WPTR if no desc were written */
+
+static void bdx_rx_alloc_skbs(struct bdx_priv *priv, struct rxf_fifo *f)
+{
+	struct sk_buff *skb;
+	struct rxf_desc *rxfd;
+	struct rx_map *dm;
+	int dno, delta, idx;
+	struct rxdb *db = priv->rxdb;
+
+	ENTER;
+	dno = bdx_rxdb_available(db) - 1;
+	while (dno > 0) {
+		skb = netdev_alloc_skb(priv->ndev, f->m.pktsz + NET_IP_ALIGN);
+		if (!skb)
+			break;
+
+		skb_reserve(skb, NET_IP_ALIGN);
+
+		idx = bdx_rxdb_alloc_elem(db);
+		dm = bdx_rxdb_addr_elem(db, idx);
+		dm->dma = dma_map_single(&priv->pdev->dev, skb->data,
+					 f->m.pktsz, DMA_FROM_DEVICE);
+		dm->skb = skb;
+		rxfd = (struct rxf_desc *)(f->m.va + f->m.wptr);
+		rxfd->info = CPU_CHIP_SWAP32(0x10003);	/* INFO=1 BC=3 */
+		rxfd->va_lo = idx;
+		rxfd->pa_lo = CPU_CHIP_SWAP32(L32_64(dm->dma));
+		rxfd->pa_hi = CPU_CHIP_SWAP32(H32_64(dm->dma));
+		rxfd->len = CPU_CHIP_SWAP32(f->m.pktsz);
+		print_rxfd(rxfd);
+
+		f->m.wptr += sizeof(struct rxf_desc);
+		delta = f->m.wptr - f->m.memsz;
+		if (unlikely(delta >= 0)) {
+			f->m.wptr = delta;
+			if (delta > 0) {
+				memcpy(f->m.va, f->m.va + f->m.memsz, delta);
+				DBG("wrapped descriptor\n");
+			}
+		}
+		dno--;
+	}
+	/*TBD: to do - delayed rxf wptr like in txd */
+	WRITE_REG(priv, f->m.reg_WPTR, f->m.wptr & TXF_WPTR_WR_PTR);
+	RET();
+}
+
+static inline void
+NETIF_RX_MUX(struct bdx_priv *priv, u32 rxd_val1, u16 rxd_vlan,
+	     struct sk_buff *skb)
+{
+	ENTER;
+	DBG("rxdd->flags.bits.vtag=%d\n", GET_RXD_VTAG(rxd_val1));
+	if (GET_RXD_VTAG(rxd_val1)) {
+		DBG("%s: vlan rcv vlan '%x' vtag '%x'\n",
+		    priv->ndev->name,
+		    GET_RXD_VLAN_ID(rxd_vlan),
+		    GET_RXD_VTAG(rxd_val1));
+		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), GET_RXD_VLAN_TCI(rxd_vlan));
+	}
+	netif_receive_skb(skb);
+}
+
+static void bdx_recycle_skb(struct bdx_priv *priv, struct rxd_desc *rxdd)
+{
+	struct rxf_desc *rxfd;
+	struct rx_map *dm;
+	struct rxf_fifo *f;
+	struct rxdb *db;
+	int delta;
+
+	ENTER;
+	DBG("priv=%p rxdd=%p\n", priv, rxdd);
+	f = &priv->rxf_fifo0;
+	db = priv->rxdb;
+	DBG("db=%p f=%p\n", db, f);
+	dm = bdx_rxdb_addr_elem(db, rxdd->va_lo);
+	DBG("dm=%p\n", dm);
+	rxfd = (struct rxf_desc *)(f->m.va + f->m.wptr);
+	rxfd->info = CPU_CHIP_SWAP32(0x10003);	/* INFO=1 BC=3 */
+	rxfd->va_lo = rxdd->va_lo;
+	rxfd->pa_lo = CPU_CHIP_SWAP32(L32_64(dm->dma));
+	rxfd->pa_hi = CPU_CHIP_SWAP32(H32_64(dm->dma));
+	rxfd->len = CPU_CHIP_SWAP32(f->m.pktsz);
+	print_rxfd(rxfd);
+
+	f->m.wptr += sizeof(struct rxf_desc);
+	delta = f->m.wptr - f->m.memsz;
+	if (unlikely(delta >= 0)) {
+		f->m.wptr = delta;
+		if (delta > 0) {
+			memcpy(f->m.va, f->m.va + f->m.memsz, delta);
+			DBG("wrapped descriptor\n");
+		}
+	}
+	RET();
+}
+
+/**
+ * bdx_rx_receive - receives full packets from RXD fifo and pass them to OS
+ * NOTE: a special treatment is given to non-continuous descriptors
+ * that start near the end, wraps around and continue at the beginning. a second
+ * part is copied right after the first, and then descriptor is interpreted as
+ * normal. fifo has an extra space to allow such operations
+ * @priv: nic's private structure
+ * @f: RXF fifo that needs skbs
+ * @budget: maximum number of packets to receive
+ */
+
+/* TBD: replace memcpy func call by explicite inline asm */
+
+static int bdx_rx_receive(struct bdx_priv *priv, struct rxd_fifo *f, int budget)
+{
+	struct net_device *ndev = priv->ndev;
+	struct sk_buff *skb, *skb2;
+	struct rxd_desc *rxdd;
+	struct rx_map *dm;
+	struct rxf_fifo *rxf_fifo;
+	int tmp_len, size;
+	int done = 0;
+	int max_done = BDX_MAX_RX_DONE;
+	struct rxdb *db = NULL;
+	/* Unmarshalled descriptor - copy of descriptor in host order */
+	u32 rxd_val1;
+	u16 len;
+	u16 rxd_vlan;
+
+	ENTER;
+	max_done = budget;
+
+	f->m.wptr = READ_REG(priv, f->m.reg_WPTR) & TXF_WPTR_WR_PTR;
+
+	size = f->m.wptr - f->m.rptr;
+	if (size < 0)
+		size = f->m.memsz + size;	/* size is negative :-) */
+
+	while (size > 0) {
+
+		rxdd = (struct rxd_desc *)(f->m.va + f->m.rptr);
+		rxd_val1 = CPU_CHIP_SWAP32(rxdd->rxd_val1);
+
+		len = CPU_CHIP_SWAP16(rxdd->len);
+
+		rxd_vlan = CPU_CHIP_SWAP16(rxdd->rxd_vlan);
+
+		print_rxdd(rxdd, rxd_val1, len, rxd_vlan);
+
+		tmp_len = GET_RXD_BC(rxd_val1) << 3;
+		BDX_ASSERT(tmp_len <= 0);
+		size -= tmp_len;
+		if (size < 0)	/* test for partially arrived descriptor */
+			break;
+
+		f->m.rptr += tmp_len;
+
+		tmp_len = f->m.rptr - f->m.memsz;
+		if (unlikely(tmp_len >= 0)) {
+			f->m.rptr = tmp_len;
+			if (tmp_len > 0) {
+				DBG("wrapped desc rptr=%d tmp_len=%d\n",
+				    f->m.rptr, tmp_len);
+				memcpy(f->m.va + f->m.memsz, f->m.va, tmp_len);
+			}
+		}
+
+		if (unlikely(GET_RXD_ERR(rxd_val1))) {
+			DBG("rxd_err = 0x%x\n", GET_RXD_ERR(rxd_val1));
+			ndev->stats.rx_errors++;
+			bdx_recycle_skb(priv, rxdd);
+			continue;
+		}
+
+		rxf_fifo = &priv->rxf_fifo0;
+		db = priv->rxdb;
+		dm = bdx_rxdb_addr_elem(db, rxdd->va_lo);
+		skb = dm->skb;
+
+		if (len < BDX_COPYBREAK &&
+		    (skb2 = netdev_alloc_skb(priv->ndev, len + NET_IP_ALIGN))) {
+			skb_reserve(skb2, NET_IP_ALIGN);
+			/*skb_put(skb2, len); */
+			dma_sync_single_for_cpu(&priv->pdev->dev, dm->dma,
+						rxf_fifo->m.pktsz,
+						DMA_FROM_DEVICE);
+			memcpy(skb2->data, skb->data, len);
+			bdx_recycle_skb(priv, rxdd);
+			skb = skb2;
+		} else {
+			dma_unmap_single(&priv->pdev->dev, dm->dma,
+					 rxf_fifo->m.pktsz, DMA_FROM_DEVICE);
+			bdx_rxdb_free_elem(db, rxdd->va_lo);
+		}
+
+		ndev->stats.rx_bytes += len;
+
+		skb_put(skb, len);
+		skb->protocol = eth_type_trans(skb, ndev);
+
+		/* Non-IP packets aren't checksum-offloaded */
+		if (GET_RXD_PKT_ID(rxd_val1) == 0)
+			skb_checksum_none_assert(skb);
+		else
+			skb->ip_summed = CHECKSUM_UNNECESSARY;
+
+		NETIF_RX_MUX(priv, rxd_val1, rxd_vlan, skb);
+
+		if (++done >= max_done)
+			break;
+	}
+
+	ndev->stats.rx_packets += done;
+
+	/* FIXME: do smth to minimize pci accesses    */
+	WRITE_REG(priv, f->m.reg_RPTR, f->m.rptr & TXF_WPTR_WR_PTR);
+
+	bdx_rx_alloc_skbs(priv, &priv->rxf_fifo0);
+
+	RET(done);
+}
+
+/*************************************************************************
+ * Debug / Temprorary Code                                               *
+ *************************************************************************/
+static void print_rxdd(struct rxd_desc *rxdd, u32 rxd_val1, u16 len,
+		       u16 rxd_vlan)
+{
+	DBG("ERROR: rxdd bc %d rxfq %d to %d type %d err %d rxp %d pkt_id %d vtag %d len %d vlan_id %d cfi %d prio %d va_lo %d va_hi %d\n",
+	    GET_RXD_BC(rxd_val1), GET_RXD_RXFQ(rxd_val1), GET_RXD_TO(rxd_val1),
+	    GET_RXD_TYPE(rxd_val1), GET_RXD_ERR(rxd_val1),
+	    GET_RXD_RXP(rxd_val1), GET_RXD_PKT_ID(rxd_val1),
+	    GET_RXD_VTAG(rxd_val1), len, GET_RXD_VLAN_ID(rxd_vlan),
+	    GET_RXD_CFI(rxd_vlan), GET_RXD_PRIO(rxd_vlan), rxdd->va_lo,
+	    rxdd->va_hi);
+}
+
+static void print_rxfd(struct rxf_desc *rxfd)
+{
+	DBG("=== RxF desc CHIP ORDER/ENDIANNESS =============\n"
+	    "info 0x%x va_lo %u pa_lo 0x%x pa_hi 0x%x len 0x%x\n",
+	    rxfd->info, rxfd->va_lo, rxfd->pa_lo, rxfd->pa_hi, rxfd->len);
+}
+
+/*
+ * TX HW/SW interaction overview
+ * ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+ * There are 2 types of TX communication channels between driver and NIC.
+ * 1) TX Free Fifo - TXF - holds ack descriptors for sent packets
+ * 2) TX Data Fifo - TXD - holds descriptors of full buffers.
+ *
+ * Currently NIC supports TSO, checksuming and gather DMA
+ * UFO and IP fragmentation is on the way
+ *
+ * RX SW Data Structures
+ * ~~~~~~~~~~~~~~~~~~~~~
+ * txdb - used to keep track of all skbs owned by SW and their dma addresses.
+ * For TX case, ownership lasts from geting packet via hard_xmit and until HW
+ * acknowledges sent by TXF descriptors.
+ * Implemented as cyclic buffer.
+ * 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.
+ *
+ * TX SW Execution Flow
+ * ~~~~~~~~~~~~~~~~~~~~
+ * OS calls driver's hard_xmit method with packet to sent.
+ * Driver creates DMA mappings, builds TXD descriptors and kicks HW
+ * by updating TXD WPTR.
+ * When packet is sent, HW write us TXF descriptor and SW frees original skb.
+ * To prevent TXD fifo overflow without reading HW registers every time,
+ * SW deploys "tx level" technique.
+ * Upon strart up, tx level is initialized to TXD fifo length.
+ * For every sent packet, SW gets its TXD descriptor sizei
+ * (from precalculated array) and substructs it from tx level.
+ * The size is also stored in txdb. When TXF ack arrives, SW fetch size of
+ * original TXD descriptor from txdb and adds it to tx level.
+ * When Tx level drops under some predefined treshhold, the driver
+ * stops the TX queue. When TX level rises above that level,
+ * the tx queue is enabled again.
+ *
+ * This technique avoids eccessive reading of RPTR and WPTR registers.
+ * As our benchmarks shows, it adds 1.5 Gbit/sec to NIS's throuput.
+ */
+
+/**
+ * __bdx_tx_db_ptr_next - helper function, increment read/write pointer + wrap
+ * @db: tx data base
+ * @pptr: read or write pointer
+ */
+static inline void __bdx_tx_db_ptr_next(struct txdb *db, struct tx_map **pptr)
+{
+	BDX_ASSERT(db == NULL || pptr == NULL);	/* sanity */
+
+	BDX_ASSERT(*pptr != db->rptr &&	/* expect either read */
+		   *pptr != db->wptr);	/* or write pointer */
+
+	BDX_ASSERT(*pptr < db->start ||	/* pointer has to be */
+		   *pptr >= db->end);	/* in range */
+
+	++*pptr;
+	if (unlikely(*pptr == db->end))
+		*pptr = db->start;
+}
+
+/**
+ * bdx_tx_db_inc_rptr - increment read pointer
+ * @db: tx data base
+ */
+static inline void bdx_tx_db_inc_rptr(struct txdb *db)
+{
+	BDX_ASSERT(db->rptr == db->wptr);	/* can't read from empty db */
+	__bdx_tx_db_ptr_next(db, &db->rptr);
+}
+
+/**
+ * bdx_tx_db_inc_wptr - increment write pointer
+ * @db: tx data base
+ */
+static inline void bdx_tx_db_inc_wptr(struct txdb *db)
+{
+	__bdx_tx_db_ptr_next(db, &db->wptr);
+	BDX_ASSERT(db->rptr == db->wptr);	/* we can not get empty db as
+						   a result of write */
+}
+
+/**
+ * bdx_tx_db_init - creates and initializes tx db
+ * @d: tx data base
+ * @sz_type: size of tx fifo
+ *
+ * Returns 0 on success, error code otherwise
+ */
+static int bdx_tx_db_init(struct txdb *d, int sz_type)
+{
+	int memsz = FIFO_SIZE * (1 << (sz_type + 1));
+
+	d->start = vmalloc(memsz);
+	if (!d->start)
+		return -ENOMEM;
+
+	/*
+	 * In order to differentiate between db is empty and db is full
+	 * states at least one element should always be empty in order to
+	 * avoid rptr == wptr which means db is empty
+	 */
+	d->size = memsz / sizeof(struct tx_map) - 1;
+	d->end = d->start + d->size + 1;	/* just after last element */
+
+	/* all dbs are created equally empty */
+	d->rptr = d->start;
+	d->wptr = d->start;
+
+	return 0;
+}
+
+/**
+ * bdx_tx_db_close - closes tx db and frees all memory
+ * @d: tx data base
+ */
+static void bdx_tx_db_close(struct txdb *d)
+{
+	BDX_ASSERT(d == NULL);
+
+	vfree(d->start);
+	d->start = NULL;
+}
+
+/*************************************************************************
+ *     Tx Engine                                                         *
+ *************************************************************************/
+
+/* sizes of tx desc (including padding if needed) as function
+ * of skb's frag number */
+static struct {
+	u16 bytes;
+	u16 qwords;		/* qword = 64 bit */
+} txd_sizes[MAX_SKB_FRAGS + 1];
+
+/**
+ * bdx_tx_map_skb - creates and stores dma mappings for skb's data blocks
+ * @priv: NIC private structure
+ * @skb: socket buffer to map
+ * @txdd: TX descriptor to use
+ *
+ * It makes dma mappings for skb's data blocks and writes them to PBL of
+ * new tx descriptor. It also stores them in the tx db, so they could be
+ * unmaped after data was sent. It is reponsibility of a caller to make
+ * sure that there is enough space in the tx db. Last element holds pointer
+ * to skb itself and marked with zero length
+ */
+static inline void
+bdx_tx_map_skb(struct bdx_priv *priv, struct sk_buff *skb,
+	       struct txd_desc *txdd)
+{
+	struct txdb *db = &priv->txdb;
+	struct pbl *pbl = &txdd->pbl[0];
+	int nr_frags = skb_shinfo(skb)->nr_frags;
+	int i;
+
+	db->wptr->len = skb_headlen(skb);
+	db->wptr->addr.dma = dma_map_single(&priv->pdev->dev, skb->data,
+					    db->wptr->len, DMA_TO_DEVICE);
+	pbl->len = CPU_CHIP_SWAP32(db->wptr->len);
+	pbl->pa_lo = CPU_CHIP_SWAP32(L32_64(db->wptr->addr.dma));
+	pbl->pa_hi = CPU_CHIP_SWAP32(H32_64(db->wptr->addr.dma));
+	DBG("=== pbl   len: 0x%x ================\n", pbl->len);
+	DBG("=== pbl pa_lo: 0x%x ================\n", pbl->pa_lo);
+	DBG("=== pbl pa_hi: 0x%x ================\n", pbl->pa_hi);
+	bdx_tx_db_inc_wptr(db);
+
+	for (i = 0; i < nr_frags; i++) {
+		const skb_frag_t *frag;
+
+		frag = &skb_shinfo(skb)->frags[i];
+		db->wptr->len = skb_frag_size(frag);
+		db->wptr->addr.dma = skb_frag_dma_map(&priv->pdev->dev, frag,
+						      0, skb_frag_size(frag),
+						      DMA_TO_DEVICE);
+
+		pbl++;
+		pbl->len = CPU_CHIP_SWAP32(db->wptr->len);
+		pbl->pa_lo = CPU_CHIP_SWAP32(L32_64(db->wptr->addr.dma));
+		pbl->pa_hi = CPU_CHIP_SWAP32(H32_64(db->wptr->addr.dma));
+		bdx_tx_db_inc_wptr(db);
+	}
+
+	/* add skb clean up info. */
+	db->wptr->len = -txd_sizes[nr_frags].bytes;
+	db->wptr->addr.skb = skb;
+	bdx_tx_db_inc_wptr(db);
+}
+
+/* init_txd_sizes - precalculate sizes of descriptors for skbs up to 16 frags
+ * number of frags is used as index to fetch correct descriptors size,
+ * instead of calculating it each time */
+static void __init init_txd_sizes(void)
+{
+	int i, lwords;
+
+	/* 7 - is number of lwords in txd with one phys buffer
+	 * 3 - is number of lwords used for every additional phys buffer */
+	for (i = 0; i < MAX_SKB_FRAGS + 1; i++) {
+		lwords = 7 + (i * 3);
+		if (lwords & 1)
+			lwords++;	/* pad it with 1 lword */
+		txd_sizes[i].qwords = lwords >> 1;
+		txd_sizes[i].bytes = lwords << 2;
+	}
+}
+
+/* bdx_tx_init - initialize all Tx related stuff.
+ * Namely, TXD and TXF fifos, database etc */
+static int bdx_tx_init(struct bdx_priv *priv)
+{
+	if (bdx_fifo_init(priv, &priv->txd_fifo0.m, priv->txd_size,
+			  regTXD_CFG0_0,
+			  regTXD_CFG1_0, regTXD_RPTR_0, regTXD_WPTR_0))
+		goto err_mem;
+	if (bdx_fifo_init(priv, &priv->txf_fifo0.m, priv->txf_size,
+			  regTXF_CFG0_0,
+			  regTXF_CFG1_0, regTXF_RPTR_0, regTXF_WPTR_0))
+		goto err_mem;
+
+	/* The TX db has to keep mappings for all packets sent (on TxD)
+	 * and not yet reclaimed (on TxF) */
+	if (bdx_tx_db_init(&priv->txdb, max(priv->txd_size, priv->txf_size)))
+		goto err_mem;
+
+	priv->tx_level = BDX_MAX_TX_LEVEL;
+#ifdef BDX_DELAY_WPTR
+	priv->tx_update_mark = priv->tx_level - 1024;
+#endif
+	return 0;
+
+err_mem:
+	netdev_err(priv->ndev, "Tx init failed\n");
+	return -ENOMEM;
+}
+
+/**
+ * bdx_tx_space - calculates available space in TX fifo
+ * @priv: NIC private structure
+ *
+ * Returns available space in TX fifo in bytes
+ */
+static inline int bdx_tx_space(struct bdx_priv *priv)
+{
+	struct txd_fifo *f = &priv->txd_fifo0;
+	int fsize;
+
+	f->m.rptr = READ_REG(priv, f->m.reg_RPTR) & TXF_WPTR_WR_PTR;
+	fsize = f->m.rptr - f->m.wptr;
+	if (fsize <= 0)
+		fsize = f->m.memsz + fsize;
+	return fsize;
+}
+
+/**
+ * bdx_tx_transmit - send packet to NIC
+ * @skb: packet to send
+ * @ndev: network device assigned to NIC
+ * Return codes:
+ * o NETDEV_TX_OK everything ok.
+ * o NETDEV_TX_BUSY Cannot transmit packet, try later
+ *   Usually a bug, means queue start/stop flow control is broken in
+ *   the driver. Note: the driver must NOT put the skb in its DMA ring.
+ */
+static netdev_tx_t bdx_tx_transmit(struct sk_buff *skb,
+				   struct net_device *ndev)
+{
+	struct bdx_priv *priv = netdev_priv(ndev);
+	struct txd_fifo *f = &priv->txd_fifo0;
+	int txd_checksum = 7;	/* full checksum */
+	int txd_lgsnd = 0;
+	int txd_vlan_id = 0;
+	int txd_vtag = 0;
+	int txd_mss = 0;
+
+	int nr_frags = skb_shinfo(skb)->nr_frags;
+	struct txd_desc *txdd;
+	int len;
+	unsigned long flags;
+
+	ENTER;
+	local_irq_save(flags);
+	spin_lock(&priv->tx_lock);
+
+	/* build tx descriptor */
+	BDX_ASSERT(f->m.wptr >= f->m.memsz);	/* started with valid wptr */
+	txdd = (struct txd_desc *)(f->m.va + f->m.wptr);
+	if (unlikely(skb->ip_summed != CHECKSUM_PARTIAL))
+		txd_checksum = 0;
+
+	if (skb_shinfo(skb)->gso_size) {
+		txd_mss = skb_shinfo(skb)->gso_size;
+		txd_lgsnd = 1;
+		DBG("skb %p skb len %d gso size = %d\n", skb, skb->len,
+		    txd_mss);
+	}
+
+	if (skb_vlan_tag_present(skb)) {
+		/*Cut VLAN ID to 12 bits */
+		txd_vlan_id = skb_vlan_tag_get(skb) & BITS_MASK(12);
+		txd_vtag = 1;
+	}
+
+	txdd->length = CPU_CHIP_SWAP16(skb->len);
+	txdd->mss = CPU_CHIP_SWAP16(txd_mss);
+	txdd->txd_val1 =
+	    CPU_CHIP_SWAP32(TXD_W1_VAL
+			    (txd_sizes[nr_frags].qwords, txd_checksum, txd_vtag,
+			     txd_lgsnd, txd_vlan_id));
+	DBG("=== TxD desc =====================\n");
+	DBG("=== w1: 0x%x ================\n", txdd->txd_val1);
+	DBG("=== w2: mss 0x%x len 0x%x\n", txdd->mss, txdd->length);
+
+	bdx_tx_map_skb(priv, skb, txdd);
+
+	/* increment TXD write pointer. In case of
+	   fifo wrapping copy reminder of the descriptor
+	   to the beginning */
+	f->m.wptr += txd_sizes[nr_frags].bytes;
+	len = f->m.wptr - f->m.memsz;
+	if (unlikely(len >= 0)) {
+		f->m.wptr = len;
+		if (len > 0) {
+			BDX_ASSERT(len > f->m.memsz);
+			memcpy(f->m.va, f->m.va + f->m.memsz, len);
+		}
+	}
+	BDX_ASSERT(f->m.wptr >= f->m.memsz);	/* finished with valid wptr */
+
+	priv->tx_level -= txd_sizes[nr_frags].bytes;
+	BDX_ASSERT(priv->tx_level <= 0 || priv->tx_level > BDX_MAX_TX_LEVEL);
+#ifdef BDX_DELAY_WPTR
+	if (priv->tx_level > priv->tx_update_mark) {
+		/* Force memory writes to complete before letting h/w
+		   know there are new descriptors to fetch.
+		   (might be needed on platforms like IA64)
+		   wmb(); */
+		WRITE_REG(priv, f->m.reg_WPTR, f->m.wptr & TXF_WPTR_WR_PTR);
+	} else {
+		if (priv->tx_noupd++ > BDX_NO_UPD_PACKETS) {
+			priv->tx_noupd = 0;
+			WRITE_REG(priv, f->m.reg_WPTR,
+				  f->m.wptr & TXF_WPTR_WR_PTR);
+		}
+	}
+#else
+	/* Force memory writes to complete before letting h/w
+	   know there are new descriptors to fetch.
+	   (might be needed on platforms like IA64)
+	   wmb(); */
+	WRITE_REG(priv, f->m.reg_WPTR, f->m.wptr & TXF_WPTR_WR_PTR);
+
+#endif
+#ifdef BDX_LLTX
+	netif_trans_update(ndev); /* NETIF_F_LLTX driver :( */
+#endif
+	ndev->stats.tx_packets++;
+	ndev->stats.tx_bytes += skb->len;
+
+	if (priv->tx_level < BDX_MIN_TX_LEVEL) {
+		DBG("%s: %s: TX Q STOP level %d\n",
+		    BDX_DRV_NAME, ndev->name, priv->tx_level);
+		netif_stop_queue(ndev);
+	}
+
+	spin_unlock_irqrestore(&priv->tx_lock, flags);
+	return NETDEV_TX_OK;
+}
+
+/**
+ * bdx_tx_cleanup - clean TXF fifo, run in the context of IRQ.
+ * @priv: bdx adapter
+ *
+ * It scans TXF fifo for descriptors, frees DMA mappings and reports to OS
+ * that those packets were sent
+ */
+static void bdx_tx_cleanup(struct bdx_priv *priv)
+{
+	struct txf_fifo *f = &priv->txf_fifo0;
+	struct txdb *db = &priv->txdb;
+	int tx_level = 0;
+
+	ENTER;
+	f->m.wptr = READ_REG(priv, f->m.reg_WPTR) & TXF_WPTR_MASK;
+	BDX_ASSERT(f->m.rptr >= f->m.memsz);	/* started with valid rptr */
+
+	while (f->m.wptr != f->m.rptr) {
+		f->m.rptr += BDX_TXF_DESC_SZ;
+		f->m.rptr &= f->m.size_mask;
+
+		/* unmap all the fragments */
+		/* first has to come tx_maps containing dma */
+		BDX_ASSERT(db->rptr->len == 0);
+		do {
+			BDX_ASSERT(db->rptr->addr.dma == 0);
+			dma_unmap_page(&priv->pdev->dev, db->rptr->addr.dma,
+				       db->rptr->len, DMA_TO_DEVICE);
+			bdx_tx_db_inc_rptr(db);
+		} while (db->rptr->len > 0);
+		tx_level -= db->rptr->len;	/* '-' koz len is negative */
+
+		/* now should come skb pointer - free it */
+		dev_consume_skb_irq(db->rptr->addr.skb);
+		bdx_tx_db_inc_rptr(db);
+	}
+
+	/* let h/w know which TXF descriptors were cleaned */
+	BDX_ASSERT((f->m.wptr & TXF_WPTR_WR_PTR) >= f->m.memsz);
+	WRITE_REG(priv, f->m.reg_RPTR, f->m.rptr & TXF_WPTR_WR_PTR);
+
+	/* We reclaimed resources, so in case the Q is stopped by xmit callback,
+	 * we resume the transmission and use tx_lock to synchronize with xmit.*/
+	spin_lock(&priv->tx_lock);
+	priv->tx_level += tx_level;
+	BDX_ASSERT(priv->tx_level <= 0 || priv->tx_level > BDX_MAX_TX_LEVEL);
+#ifdef BDX_DELAY_WPTR
+	if (priv->tx_noupd) {
+		priv->tx_noupd = 0;
+		WRITE_REG(priv, priv->txd_fifo0.m.reg_WPTR,
+			  priv->txd_fifo0.m.wptr & TXF_WPTR_WR_PTR);
+	}
+#endif
+
+	if (unlikely(netif_queue_stopped(priv->ndev) &&
+		     netif_carrier_ok(priv->ndev) &&
+		     (priv->tx_level >= BDX_MIN_TX_LEVEL))) {
+		DBG("%s: %s: TX Q WAKE level %d\n",
+		    BDX_DRV_NAME, priv->ndev->name, priv->tx_level);
+		netif_wake_queue(priv->ndev);
+	}
+	spin_unlock(&priv->tx_lock);
+}
+
+/**
+ * bdx_tx_free_skbs - frees all skbs from TXD fifo.
+ * @priv: NIC private structure
+ *
+ * It gets called when OS stops this dev, eg upon "ifconfig down" or rmmod
+ */
+static void bdx_tx_free_skbs(struct bdx_priv *priv)
+{
+	struct txdb *db = &priv->txdb;
+
+	ENTER;
+	while (db->rptr != db->wptr) {
+		if (likely(db->rptr->len))
+			dma_unmap_page(&priv->pdev->dev, db->rptr->addr.dma,
+				       db->rptr->len, DMA_TO_DEVICE);
+		else
+			dev_kfree_skb(db->rptr->addr.skb);
+		bdx_tx_db_inc_rptr(db);
+	}
+	RET();
+}
+
+/* bdx_tx_free - frees all Tx resources */
+static void bdx_tx_free(struct bdx_priv *priv)
+{
+	ENTER;
+	bdx_tx_free_skbs(priv);
+	bdx_fifo_free(priv, &priv->txd_fifo0.m);
+	bdx_fifo_free(priv, &priv->txf_fifo0.m);
+	bdx_tx_db_close(&priv->txdb);
+}
+
+/**
+ * bdx_tx_push_desc - push descriptor to TxD fifo
+ * @priv: NIC private structure
+ * @data: desc's data
+ * @size: desc's size
+ *
+ * Pushes desc to TxD fifo and overlaps it if needed.
+ * NOTE: this func does not check for available space. this is responsibility
+ *    of the caller. Neither does it check that data size is smaller than
+ *    fifo size.
+ */
+static void bdx_tx_push_desc(struct bdx_priv *priv, void *data, int size)
+{
+	struct txd_fifo *f = &priv->txd_fifo0;
+	int i = f->m.memsz - f->m.wptr;
+
+	if (size == 0)
+		return;
+
+	if (i > size) {
+		memcpy(f->m.va + f->m.wptr, data, size);
+		f->m.wptr += size;
+	} else {
+		memcpy(f->m.va + f->m.wptr, data, i);
+		f->m.wptr = size - i;
+		memcpy(f->m.va, data + i, f->m.wptr);
+	}
+	WRITE_REG(priv, f->m.reg_WPTR, f->m.wptr & TXF_WPTR_WR_PTR);
+}
+
+/**
+ * bdx_tx_push_desc_safe - push descriptor to TxD fifo in a safe way
+ * @priv: NIC private structure
+ * @data: desc's data
+ * @size: desc's size
+ *
+ * NOTE: this func does check for available space and, if necessary, waits for
+ *   NIC to read existing data before writing new one.
+ */
+static void bdx_tx_push_desc_safe(struct bdx_priv *priv, void *data, int size)
+{
+	int timer = 0;
+	ENTER;
+
+	while (size > 0) {
+		/* we substruct 8 because when fifo is full rptr == wptr
+		   which also means that fifo is empty, we can understand
+		   the difference, but could hw do the same ??? :) */
+		int avail = bdx_tx_space(priv) - 8;
+		if (avail <= 0) {
+			if (timer++ > 300) {	/* prevent endless loop */
+				DBG("timeout while writing desc to TxD fifo\n");
+				break;
+			}
+			udelay(50);	/* give hw a chance to clean fifo */
+			continue;
+		}
+		avail = min(avail, size);
+		DBG("about to push  %d bytes starting %p size %d\n", avail,
+		    data, size);
+		bdx_tx_push_desc(priv, data, avail);
+		size -= avail;
+		data += avail;
+	}
+	RET();
+}
+
+static const struct net_device_ops bdx_netdev_ops = {
+	.ndo_open		= bdx_open,
+	.ndo_stop		= bdx_close,
+	.ndo_start_xmit		= bdx_tx_transmit,
+	.ndo_validate_addr	= eth_validate_addr,
+	.ndo_siocdevprivate	= bdx_siocdevprivate,
+	.ndo_set_rx_mode	= bdx_setmulti,
+	.ndo_change_mtu		= bdx_change_mtu,
+	.ndo_set_mac_address	= bdx_set_mac,
+	.ndo_vlan_rx_add_vid	= bdx_vlan_rx_add_vid,
+	.ndo_vlan_rx_kill_vid	= bdx_vlan_rx_kill_vid,
+};
+
+/**
+ * bdx_probe - Device Initialization Routine
+ * @pdev: PCI device information struct
+ * @ent: entry in bdx_pci_tbl
+ *
+ * Returns 0 on success, negative on failure
+ *
+ * bdx_probe initializes an adapter identified by a pci_dev structure.
+ * The OS initialization, configuring of the adapter private structure,
+ * and a hardware reset occur.
+ *
+ * functions and their order used as explained in
+ * /usr/src/linux/Documentation/DMA-{API,mapping}.txt
+ *
+ */
+
+/* TBD: netif_msg should be checked and implemented. I disable it for now */
+static int
+bdx_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
+{
+	struct net_device *ndev;
+	struct bdx_priv *priv;
+	unsigned long pciaddr;
+	u32 regionSize;
+	struct pci_nic *nic;
+	int err, port;
+
+	ENTER;
+
+	nic = vmalloc(sizeof(*nic));
+	if (!nic)
+		RET(-ENOMEM);
+
+    /************** pci *****************/
+	err = pci_enable_device(pdev);
+	if (err)			/* it triggers interrupt, dunno why. */
+		goto err_pci;		/* it's not a problem though */
+
+	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
+	if (err) {
+		pr_err("No usable DMA configuration, aborting\n");
+		goto err_dma;
+	}
+
+	err = pci_request_regions(pdev, BDX_DRV_NAME);
+	if (err)
+		goto err_dma;
+
+	pci_set_master(pdev);
+
+	pciaddr = pci_resource_start(pdev, 0);
+	if (!pciaddr) {
+		err = -EIO;
+		pr_err("no MMIO resource\n");
+		goto err_out_res;
+	}
+	regionSize = pci_resource_len(pdev, 0);
+	if (regionSize < BDX_REGS_SIZE) {
+		err = -EIO;
+		pr_err("MMIO resource (%x) too small\n", regionSize);
+		goto err_out_res;
+	}
+
+	nic->regs = ioremap(pciaddr, regionSize);
+	if (!nic->regs) {
+		err = -EIO;
+		pr_err("ioremap failed\n");
+		goto err_out_res;
+	}
+
+	if (pdev->irq < 2) {
+		err = -EIO;
+		pr_err("invalid irq (%d)\n", pdev->irq);
+		goto err_out_iomap;
+	}
+	pci_set_drvdata(pdev, nic);
+
+	if (pdev->device == 0x3014)
+		nic->port_num = 2;
+	else
+		nic->port_num = 1;
+
+	print_hw_id(pdev);
+
+	bdx_hw_reset_direct(nic->regs);
+
+	nic->irq_type = IRQ_INTX;
+#ifdef BDX_MSI
+	if ((readl(nic->regs + FPGA_VER) & 0xFFF) >= 378) {
+		err = pci_enable_msi(pdev);
+		if (err)
+			pr_err("Can't enable msi. error is %d\n", err);
+		else
+			nic->irq_type = IRQ_MSI;
+	} else
+		DBG("HW does not support MSI\n");
+#endif
+
+    /************** netdev **************/
+	for (port = 0; port < nic->port_num; port++) {
+		ndev = alloc_etherdev(sizeof(struct bdx_priv));
+		if (!ndev) {
+			err = -ENOMEM;
+			goto err_out_iomap;
+		}
+
+		ndev->netdev_ops = &bdx_netdev_ops;
+		ndev->tx_queue_len = BDX_NDEV_TXQ_LEN;
+
+		bdx_set_ethtool_ops(ndev);	/* ethtool interface */
+
+		/* these fields are used for info purposes only
+		 * so we can have them same for all ports of the board */
+		ndev->if_port = port;
+		ndev->features = NETIF_F_IP_CSUM | NETIF_F_SG | NETIF_F_TSO |
+		    NETIF_F_HW_VLAN_CTAG_TX | NETIF_F_HW_VLAN_CTAG_RX |
+		    NETIF_F_HW_VLAN_CTAG_FILTER | NETIF_F_RXCSUM |
+		    NETIF_F_HIGHDMA;
+
+		ndev->hw_features = NETIF_F_IP_CSUM | NETIF_F_SG |
+			NETIF_F_TSO | NETIF_F_HW_VLAN_CTAG_TX;
+
+	/************** priv ****************/
+		priv = nic->priv[port] = netdev_priv(ndev);
+
+		priv->pBdxRegs = nic->regs + port * 0x8000;
+		priv->port = port;
+		priv->pdev = pdev;
+		priv->ndev = ndev;
+		priv->nic = nic;
+		priv->msg_enable = BDX_DEF_MSG_ENABLE;
+
+		netif_napi_add(ndev, &priv->napi, bdx_poll);
+
+		if ((readl(nic->regs + FPGA_VER) & 0xFFF) == 308) {
+			DBG("HW statistics not supported\n");
+			priv->stats_flag = 0;
+		} else {
+			priv->stats_flag = 1;
+		}
+
+		/* Initialize fifo sizes. */
+		priv->txd_size = 2;
+		priv->txf_size = 2;
+		priv->rxd_size = 2;
+		priv->rxf_size = 3;
+
+		/* Initialize the initial coalescing registers. */
+		priv->rdintcm = INT_REG_VAL(0x20, 1, 4, 12);
+		priv->tdintcm = INT_REG_VAL(0x20, 1, 0, 12);
+
+		/* ndev->xmit_lock spinlock is not used.
+		 * Private priv->tx_lock is used for synchronization
+		 * between transmit and TX irq cleanup.  In addition
+		 * set multicast list callback has to use priv->tx_lock.
+		 */
+#ifdef BDX_LLTX
+		ndev->features |= NETIF_F_LLTX;
+#endif
+		/* MTU range: 60 - 16384 */
+		ndev->min_mtu = ETH_ZLEN;
+		ndev->max_mtu = BDX_MAX_MTU;
+
+		spin_lock_init(&priv->tx_lock);
+
+		/*bdx_hw_reset(priv); */
+		if (bdx_read_mac(priv)) {
+			pr_err("load MAC address failed\n");
+			err = -EFAULT;
+			goto err_out_iomap;
+		}
+		SET_NETDEV_DEV(ndev, &pdev->dev);
+		err = register_netdev(ndev);
+		if (err) {
+			pr_err("register_netdev failed\n");
+			goto err_out_free;
+		}
+		netif_carrier_off(ndev);
+		netif_stop_queue(ndev);
+
+		print_eth_id(ndev);
+	}
+	RET(0);
+
+err_out_free:
+	free_netdev(ndev);
+err_out_iomap:
+	iounmap(nic->regs);
+err_out_res:
+	pci_release_regions(pdev);
+err_dma:
+	pci_disable_device(pdev);
+err_pci:
+	vfree(nic);
+
+	RET(err);
+}
+
+/****************** Ethtool interface *********************/
+/* get strings for statistics counters */
+static const char
+ bdx_stat_names[][ETH_GSTRING_LEN] = {
+	"InUCast",		/* 0x7200 */
+	"InMCast",		/* 0x7210 */
+	"InBCast",		/* 0x7220 */
+	"InPkts",		/* 0x7230 */
+	"InErrors",		/* 0x7240 */
+	"InDropped",		/* 0x7250 */
+	"FrameTooLong",		/* 0x7260 */
+	"FrameSequenceErrors",	/* 0x7270 */
+	"InVLAN",		/* 0x7280 */
+	"InDroppedDFE",		/* 0x7290 */
+	"InDroppedIntFull",	/* 0x72A0 */
+	"InFrameAlignErrors",	/* 0x72B0 */
+
+	/* 0x72C0-0x72E0 RSRV */
+
+	"OutUCast",		/* 0x72F0 */
+	"OutMCast",		/* 0x7300 */
+	"OutBCast",		/* 0x7310 */
+	"OutPkts",		/* 0x7320 */
+
+	/* 0x7330-0x7360 RSRV */
+
+	"OutVLAN",		/* 0x7370 */
+	"InUCastOctects",	/* 0x7380 */
+	"OutUCastOctects",	/* 0x7390 */
+
+	/* 0x73A0-0x73B0 RSRV */
+
+	"InBCastOctects",	/* 0x73C0 */
+	"OutBCastOctects",	/* 0x73D0 */
+	"InOctects",		/* 0x73E0 */
+	"OutOctects",		/* 0x73F0 */
+};
+
+/*
+ * bdx_get_link_ksettings - get device-specific settings
+ * @netdev
+ * @ecmd
+ */
+static int bdx_get_link_ksettings(struct net_device *netdev,
+				  struct ethtool_link_ksettings *ecmd)
+{
+	ethtool_link_ksettings_zero_link_mode(ecmd, supported);
+	ethtool_link_ksettings_add_link_mode(ecmd, supported,
+					     10000baseT_Full);
+	ethtool_link_ksettings_add_link_mode(ecmd, supported, FIBRE);
+	ethtool_link_ksettings_zero_link_mode(ecmd, advertising);
+	ethtool_link_ksettings_add_link_mode(ecmd, advertising,
+					     10000baseT_Full);
+	ethtool_link_ksettings_add_link_mode(ecmd, advertising, FIBRE);
+
+	ecmd->base.speed = SPEED_10000;
+	ecmd->base.duplex = DUPLEX_FULL;
+	ecmd->base.port = PORT_FIBRE;
+	ecmd->base.autoneg = AUTONEG_DISABLE;
+
+	return 0;
+}
+
+/*
+ * bdx_get_drvinfo - report driver information
+ * @netdev
+ * @drvinfo
+ */
+static void
+bdx_get_drvinfo(struct net_device *netdev, struct ethtool_drvinfo *drvinfo)
+{
+	struct bdx_priv *priv = netdev_priv(netdev);
+
+	strscpy(drvinfo->driver, BDX_DRV_NAME, sizeof(drvinfo->driver));
+	strscpy(drvinfo->version, BDX_DRV_VERSION, sizeof(drvinfo->version));
+	strscpy(drvinfo->fw_version, "N/A", sizeof(drvinfo->fw_version));
+	strscpy(drvinfo->bus_info, pci_name(priv->pdev),
+		sizeof(drvinfo->bus_info));
+}
+
+/*
+ * bdx_get_coalesce - get interrupt coalescing parameters
+ * @netdev
+ * @ecoal
+ */
+static int bdx_get_coalesce(struct net_device *netdev,
+			    struct ethtool_coalesce *ecoal,
+			    struct kernel_ethtool_coalesce *kernel_coal,
+			    struct netlink_ext_ack *extack)
+{
+	u32 rdintcm;
+	u32 tdintcm;
+	struct bdx_priv *priv = netdev_priv(netdev);
+
+	rdintcm = priv->rdintcm;
+	tdintcm = priv->tdintcm;
+
+	/* PCK_TH measures in multiples of FIFO bytes
+	   We translate to packets */
+	ecoal->rx_coalesce_usecs = GET_INT_COAL(rdintcm) * INT_COAL_MULT;
+	ecoal->rx_max_coalesced_frames =
+	    ((GET_PCK_TH(rdintcm) * PCK_TH_MULT) / sizeof(struct rxf_desc));
+
+	ecoal->tx_coalesce_usecs = GET_INT_COAL(tdintcm) * INT_COAL_MULT;
+	ecoal->tx_max_coalesced_frames =
+	    ((GET_PCK_TH(tdintcm) * PCK_TH_MULT) / BDX_TXF_DESC_SZ);
+
+	/* adaptive parameters ignored */
+	return 0;
+}
+
+/*
+ * bdx_set_coalesce - set interrupt coalescing parameters
+ * @netdev
+ * @ecoal
+ */
+static int bdx_set_coalesce(struct net_device *netdev,
+			    struct ethtool_coalesce *ecoal,
+			    struct kernel_ethtool_coalesce *kernel_coal,
+			    struct netlink_ext_ack *extack)
+{
+	u32 rdintcm;
+	u32 tdintcm;
+	struct bdx_priv *priv = netdev_priv(netdev);
+	int rx_coal;
+	int tx_coal;
+	int rx_max_coal;
+	int tx_max_coal;
+
+	/* Check for valid input */
+	rx_coal = ecoal->rx_coalesce_usecs / INT_COAL_MULT;
+	tx_coal = ecoal->tx_coalesce_usecs / INT_COAL_MULT;
+	rx_max_coal = ecoal->rx_max_coalesced_frames;
+	tx_max_coal = ecoal->tx_max_coalesced_frames;
+
+	/* Translate from packets to multiples of FIFO bytes */
+	rx_max_coal =
+	    (((rx_max_coal * sizeof(struct rxf_desc)) + PCK_TH_MULT - 1)
+	     / PCK_TH_MULT);
+	tx_max_coal =
+	    (((tx_max_coal * BDX_TXF_DESC_SZ) + PCK_TH_MULT - 1)
+	     / PCK_TH_MULT);
+
+	if ((rx_coal > 0x7FFF) || (tx_coal > 0x7FFF) ||
+	    (rx_max_coal > 0xF) || (tx_max_coal > 0xF))
+		return -EINVAL;
+
+	rdintcm = INT_REG_VAL(rx_coal, GET_INT_COAL_RC(priv->rdintcm),
+			      GET_RXF_TH(priv->rdintcm), rx_max_coal);
+	tdintcm = INT_REG_VAL(tx_coal, GET_INT_COAL_RC(priv->tdintcm), 0,
+			      tx_max_coal);
+
+	priv->rdintcm = rdintcm;
+	priv->tdintcm = tdintcm;
+
+	WRITE_REG(priv, regRDINTCM0, rdintcm);
+	WRITE_REG(priv, regTDINTCM0, tdintcm);
+
+	return 0;
+}
+
+/* Convert RX fifo size to number of pending packets */
+static inline int bdx_rx_fifo_size_to_packets(int rx_size)
+{
+	return (FIFO_SIZE * (1 << rx_size)) / sizeof(struct rxf_desc);
+}
+
+/* Convert TX fifo size to number of pending packets */
+static inline int bdx_tx_fifo_size_to_packets(int tx_size)
+{
+	return (FIFO_SIZE * (1 << tx_size)) / BDX_TXF_DESC_SZ;
+}
+
+/*
+ * bdx_get_ringparam - report ring sizes
+ * @netdev
+ * @ring
+ * @kernel_ring
+ * @extack
+ */
+static void
+bdx_get_ringparam(struct net_device *netdev, struct ethtool_ringparam *ring,
+		  struct kernel_ethtool_ringparam *kernel_ring,
+		  struct netlink_ext_ack *extack)
+{
+	struct bdx_priv *priv = netdev_priv(netdev);
+
+	/*max_pending - the maximum-sized FIFO we allow */
+	ring->rx_max_pending = bdx_rx_fifo_size_to_packets(3);
+	ring->tx_max_pending = bdx_tx_fifo_size_to_packets(3);
+	ring->rx_pending = bdx_rx_fifo_size_to_packets(priv->rxf_size);
+	ring->tx_pending = bdx_tx_fifo_size_to_packets(priv->txd_size);
+}
+
+/*
+ * bdx_set_ringparam - set ring sizes
+ * @netdev
+ * @ring
+ * @kernel_ring
+ * @extack
+ */
+static int
+bdx_set_ringparam(struct net_device *netdev, struct ethtool_ringparam *ring,
+		  struct kernel_ethtool_ringparam *kernel_ring,
+		  struct netlink_ext_ack *extack)
+{
+	struct bdx_priv *priv = netdev_priv(netdev);
+	int rx_size = 0;
+	int tx_size = 0;
+
+	for (; rx_size < 4; rx_size++) {
+		if (bdx_rx_fifo_size_to_packets(rx_size) >= ring->rx_pending)
+			break;
+	}
+	if (rx_size == 4)
+		rx_size = 3;
+
+	for (; tx_size < 4; tx_size++) {
+		if (bdx_tx_fifo_size_to_packets(tx_size) >= ring->tx_pending)
+			break;
+	}
+	if (tx_size == 4)
+		tx_size = 3;
+
+	/*Is there anything to do? */
+	if ((rx_size == priv->rxf_size) &&
+	    (tx_size == priv->txd_size))
+		return 0;
+
+	priv->rxf_size = rx_size;
+	if (rx_size > 1)
+		priv->rxd_size = rx_size - 1;
+	else
+		priv->rxd_size = rx_size;
+
+	priv->txf_size = priv->txd_size = tx_size;
+
+	if (netif_running(netdev)) {
+		bdx_close(netdev);
+		bdx_open(netdev);
+	}
+	return 0;
+}
+
+/*
+ * bdx_get_strings - return a set of strings that describe the requested objects
+ * @netdev
+ * @data
+ */
+static void bdx_get_strings(struct net_device *netdev, u32 stringset, u8 *data)
+{
+	switch (stringset) {
+	case ETH_SS_STATS:
+		memcpy(data, *bdx_stat_names, sizeof(bdx_stat_names));
+		break;
+	}
+}
+
+/*
+ * bdx_get_sset_count - return number of statistics or tests
+ * @netdev
+ */
+static int bdx_get_sset_count(struct net_device *netdev, int stringset)
+{
+	struct bdx_priv *priv = netdev_priv(netdev);
+
+	switch (stringset) {
+	case ETH_SS_STATS:
+		BDX_ASSERT(ARRAY_SIZE(bdx_stat_names)
+			   != sizeof(struct bdx_stats) / sizeof(u64));
+		return (priv->stats_flag) ? ARRAY_SIZE(bdx_stat_names)	: 0;
+	}
+
+	return -EINVAL;
+}
+
+/*
+ * bdx_get_ethtool_stats - return device's hardware L2 statistics
+ * @netdev
+ * @stats
+ * @data
+ */
+static void bdx_get_ethtool_stats(struct net_device *netdev,
+				  struct ethtool_stats *stats, u64 *data)
+{
+	struct bdx_priv *priv = netdev_priv(netdev);
+
+	if (priv->stats_flag) {
+
+		/* Update stats from HW */
+		bdx_update_stats(priv);
+
+		/* Copy data to user buffer */
+		memcpy(data, &priv->hw_stats, sizeof(priv->hw_stats));
+	}
+}
+
+/*
+ * bdx_set_ethtool_ops - ethtool interface implementation
+ * @netdev
+ */
+static void bdx_set_ethtool_ops(struct net_device *netdev)
+{
+	static const struct ethtool_ops bdx_ethtool_ops = {
+		.supported_coalesce_params = ETHTOOL_COALESCE_USECS |
+					     ETHTOOL_COALESCE_MAX_FRAMES,
+		.get_drvinfo = bdx_get_drvinfo,
+		.get_link = ethtool_op_get_link,
+		.get_coalesce = bdx_get_coalesce,
+		.set_coalesce = bdx_set_coalesce,
+		.get_ringparam = bdx_get_ringparam,
+		.set_ringparam = bdx_set_ringparam,
+		.get_strings = bdx_get_strings,
+		.get_sset_count = bdx_get_sset_count,
+		.get_ethtool_stats = bdx_get_ethtool_stats,
+		.get_link_ksettings = bdx_get_link_ksettings,
+	};
+
+	netdev->ethtool_ops = &bdx_ethtool_ops;
+}
+
+/**
+ * bdx_remove - Device Removal Routine
+ * @pdev: PCI device information struct
+ *
+ * bdx_remove is called by the PCI subsystem to alert the driver
+ * that it should release a PCI device.  The could be caused by a
+ * Hot-Plug event, or because the driver is going to be removed from
+ * memory.
+ **/
+static void bdx_remove(struct pci_dev *pdev)
+{
+	struct pci_nic *nic = pci_get_drvdata(pdev);
+	struct net_device *ndev;
+	int port;
+
+	for (port = 0; port < nic->port_num; port++) {
+		ndev = nic->priv[port]->ndev;
+		unregister_netdev(ndev);
+		free_netdev(ndev);
+	}
+
+	/*bdx_hw_reset_direct(nic->regs); */
+#ifdef BDX_MSI
+	if (nic->irq_type == IRQ_MSI)
+		pci_disable_msi(pdev);
+#endif
+
+	iounmap(nic->regs);
+	pci_release_regions(pdev);
+	pci_disable_device(pdev);
+	vfree(nic);
+
+	RET();
+}
+
+static struct pci_driver bdx_pci_driver = {
+	.name = BDX_DRV_NAME,
+	.id_table = bdx_pci_tbl,
+	.probe = bdx_probe,
+	.remove = bdx_remove,
+};
+
+/*
+ * print_driver_id - print parameters of the driver build
+ */
+static void __init print_driver_id(void)
+{
+	pr_info("%s, %s\n", BDX_DRV_DESC, BDX_DRV_VERSION);
+	pr_info("Options: hw_csum %s\n", BDX_MSI_STRING);
+}
+
+static int __init bdx_module_init(void)
+{
+	ENTER;
+	init_txd_sizes();
+	print_driver_id();
+	RET(pci_register_driver(&bdx_pci_driver));
+}
+
+module_init(bdx_module_init);
+
+static void __exit bdx_module_exit(void)
+{
+	ENTER;
+	pci_unregister_driver(&bdx_pci_driver);
+	RET();
+}
+
+module_exit(bdx_module_exit);
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR(DRIVER_AUTHOR);
+MODULE_DESCRIPTION(BDX_DRV_DESC);
+MODULE_FIRMWARE("tehuti/bdx.bin");
diff --git a/drivers/net/ethernet/tehuti/tehuti.h b/drivers/net/ethernet/tehuti/tehuti.h
new file mode 100644
index 000000000..909e7296c
--- /dev/null
+++ b/drivers/net/ethernet/tehuti/tehuti.h
@@ -0,0 +1,560 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * Tehuti Networks(R) Network Driver
+ * Copyright (C) 2007 Tehuti Networks Ltd. All rights reserved
+ */
+
+#ifndef _TEHUTI_H
+#define _TEHUTI_H
+
+#include <linux/module.h>
+#include <linux/kernel.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/pci.h>
+#include <linux/delay.h>
+#include <linux/ethtool.h>
+#include <linux/mii.h>
+#include <linux/crc32.h>
+#include <linux/uaccess.h>
+#include <linux/in.h>
+#include <linux/ip.h>
+#include <linux/tcp.h>
+#include <linux/sched.h>
+#include <linux/tty.h>
+#include <linux/if_vlan.h>
+#include <linux/interrupt.h>
+#include <linux/vmalloc.h>
+#include <linux/firmware.h>
+#include <asm/byteorder.h>
+#include <linux/dma-mapping.h>
+#include <linux/slab.h>
+
+/* Compile Time Switches */
+/* start */
+#define BDX_TSO
+#define BDX_LLTX
+#define BDX_DELAY_WPTR
+/* #define BDX_MSI */
+/* end */
+
+#if !defined CONFIG_PCI_MSI
+#   undef BDX_MSI
+#endif
+
+#define BDX_DEF_MSG_ENABLE	(NETIF_MSG_DRV          | \
+				NETIF_MSG_PROBE        | \
+				NETIF_MSG_LINK)
+
+/* ioctl ops */
+#define BDX_OP_READ  1
+#define BDX_OP_WRITE 2
+
+/* RX copy break size */
+#define BDX_COPYBREAK    257
+
+#define DRIVER_AUTHOR     "Tehuti Networks(R)"
+#define BDX_DRV_DESC      "Tehuti Networks(R) Network Driver"
+#define BDX_DRV_NAME      "tehuti"
+#define BDX_NIC_NAME      "Tehuti 10 Giga TOE SmartNIC"
+#define BDX_NIC2PORT_NAME "Tehuti 2-Port 10 Giga TOE SmartNIC"
+#define BDX_DRV_VERSION   "7.29.3"
+
+#ifdef BDX_MSI
+#    define BDX_MSI_STRING "msi "
+#else
+#    define BDX_MSI_STRING ""
+#endif
+
+/* netdev tx queue len for Luxor. default value is, btw, 1000
+ * ifcontig eth1 txqueuelen 3000 - to change it at runtime */
+#define BDX_NDEV_TXQ_LEN 3000
+
+/* Max MTU for Jumbo Frame mode, per tehutinetworks.net Features FAQ is 16k */
+#define BDX_MAX_MTU	(16 * 1024)
+
+#define FIFO_SIZE  4096
+#define FIFO_EXTRA_SPACE            1024
+
+#if BITS_PER_LONG == 64
+#    define H32_64(x)  (u32) ((u64)(x) >> 32)
+#    define L32_64(x)  (u32) ((u64)(x) & 0xffffffff)
+#elif BITS_PER_LONG == 32
+#    define H32_64(x)  0
+#    define L32_64(x)  ((u32) (x))
+#else				/* BITS_PER_LONG == ?? */
+#    error BITS_PER_LONG is undefined. Must be 64 or 32
+#endif				/* BITS_PER_LONG */
+
+#ifdef __BIG_ENDIAN
+#   define CPU_CHIP_SWAP32(x) swab32(x)
+#   define CPU_CHIP_SWAP16(x) swab16(x)
+#else
+#   define CPU_CHIP_SWAP32(x) (x)
+#   define CPU_CHIP_SWAP16(x) (x)
+#endif
+
+#define READ_REG(pp, reg)         readl(pp->pBdxRegs + reg)
+#define WRITE_REG(pp, reg, val)   writel(val, pp->pBdxRegs + reg)
+
+#ifndef NET_IP_ALIGN
+#   define NET_IP_ALIGN 2
+#endif
+
+#ifndef NETDEV_TX_OK
+#   define NETDEV_TX_OK 0
+#endif
+
+#define LUXOR_MAX_PORT     2
+#define BDX_MAX_RX_DONE    150
+#define BDX_TXF_DESC_SZ    16
+#define BDX_MAX_TX_LEVEL   (priv->txd_fifo0.m.memsz - 16)
+#define BDX_MIN_TX_LEVEL   256
+#define BDX_NO_UPD_PACKETS 40
+
+struct pci_nic {
+	int port_num;
+	void __iomem *regs;
+	int irq_type;
+	struct bdx_priv *priv[LUXOR_MAX_PORT];
+};
+
+enum { IRQ_INTX, IRQ_MSI, IRQ_MSIX };
+
+#define PCK_TH_MULT   128
+#define INT_COAL_MULT 2
+
+#define BITS_MASK(nbits)			((1<<nbits)-1)
+#define GET_BITS_SHIFT(x, nbits, nshift)	(((x)>>nshift)&BITS_MASK(nbits))
+#define BITS_SHIFT_MASK(nbits, nshift)		(BITS_MASK(nbits)<<nshift)
+#define BITS_SHIFT_VAL(x, nbits, nshift)	(((x)&BITS_MASK(nbits))<<nshift)
+#define BITS_SHIFT_CLEAR(x, nbits, nshift)	\
+	((x)&(~BITS_SHIFT_MASK(nbits, nshift)))
+
+#define GET_INT_COAL(x)				GET_BITS_SHIFT(x, 15, 0)
+#define GET_INT_COAL_RC(x)			GET_BITS_SHIFT(x, 1, 15)
+#define GET_RXF_TH(x)				GET_BITS_SHIFT(x, 4, 16)
+#define GET_PCK_TH(x)				GET_BITS_SHIFT(x, 4, 20)
+
+#define INT_REG_VAL(coal, coal_rc, rxf_th, pck_th)	\
+	((coal)|((coal_rc)<<15)|((rxf_th)<<16)|((pck_th)<<20))
+
+struct fifo {
+	dma_addr_t da;		/* physical address of fifo (used by HW) */
+	char *va;		/* virtual address of fifo (used by SW) */
+	u32 rptr, wptr;		/* cached values of RPTR and WPTR registers,
+				   they're 32 bits on both 32 and 64 archs */
+	u16 reg_CFG0, reg_CFG1;
+	u16 reg_RPTR, reg_WPTR;
+	u16 memsz;		/* memory size allocated for fifo */
+	u16 size_mask;
+	u16 pktsz;		/* skb packet size to allocate */
+	u16 rcvno;		/* number of buffers that come from this RXF */
+};
+
+struct txf_fifo {
+	struct fifo m;		/* minimal set of variables used by all fifos */
+};
+
+struct txd_fifo {
+	struct fifo m;		/* minimal set of variables used by all fifos */
+};
+
+struct rxf_fifo {
+	struct fifo m;		/* minimal set of variables used by all fifos */
+};
+
+struct rxd_fifo {
+	struct fifo m;		/* minimal set of variables used by all fifos */
+};
+
+struct rx_map {
+	u64 dma;
+	struct sk_buff *skb;
+};
+
+struct rxdb {
+	int *stack;
+	struct rx_map *elems;
+	int nelem;
+	int top;
+};
+
+union bdx_dma_addr {
+	dma_addr_t dma;
+	struct sk_buff *skb;
+};
+
+/* Entry in the db.
+ * if len == 0 addr is dma
+ * if len != 0 addr is skb */
+struct tx_map {
+	union bdx_dma_addr addr;
+	int len;
+};
+
+/* tx database - implemented as circular fifo buffer*/
+struct txdb {
+	struct tx_map *start;	/* points to the first element */
+	struct tx_map *end;	/* points just AFTER the last element */
+	struct tx_map *rptr;	/* points to the next element to read */
+	struct tx_map *wptr;	/* points to the next element to write */
+	int size;		/* number of elements in the db */
+};
+
+/*Internal stats structure*/
+struct bdx_stats {
+	u64 InUCast;			/* 0x7200 */
+	u64 InMCast;			/* 0x7210 */
+	u64 InBCast;			/* 0x7220 */
+	u64 InPkts;			/* 0x7230 */
+	u64 InErrors;			/* 0x7240 */
+	u64 InDropped;			/* 0x7250 */
+	u64 FrameTooLong;		/* 0x7260 */
+	u64 FrameSequenceErrors;	/* 0x7270 */
+	u64 InVLAN;			/* 0x7280 */
+	u64 InDroppedDFE;		/* 0x7290 */
+	u64 InDroppedIntFull;		/* 0x72A0 */
+	u64 InFrameAlignErrors;		/* 0x72B0 */
+
+	/* 0x72C0-0x72E0 RSRV */
+
+	u64 OutUCast;			/* 0x72F0 */
+	u64 OutMCast;			/* 0x7300 */
+	u64 OutBCast;			/* 0x7310 */
+	u64 OutPkts;			/* 0x7320 */
+
+	/* 0x7330-0x7360 RSRV */
+
+	u64 OutVLAN;			/* 0x7370 */
+	u64 InUCastOctects;		/* 0x7380 */
+	u64 OutUCastOctects;		/* 0x7390 */
+
+	/* 0x73A0-0x73B0 RSRV */
+
+	u64 InBCastOctects;		/* 0x73C0 */
+	u64 OutBCastOctects;		/* 0x73D0 */
+	u64 InOctects;			/* 0x73E0 */
+	u64 OutOctects;			/* 0x73F0 */
+};
+
+struct bdx_priv {
+	void __iomem *pBdxRegs;
+	struct net_device *ndev;
+
+	struct napi_struct napi;
+
+	/* RX FIFOs: 1 for data (full) descs, and 2 for free descs */
+	struct rxd_fifo rxd_fifo0;
+	struct rxf_fifo rxf_fifo0;
+	struct rxdb *rxdb;	/* rx dbs to store skb pointers */
+	int napi_stop;
+
+	/* Tx FIFOs: 1 for data desc, 1 for empty (acks) desc */
+	struct txd_fifo txd_fifo0;
+	struct txf_fifo txf_fifo0;
+
+	struct txdb txdb;
+	int tx_level;
+#ifdef BDX_DELAY_WPTR
+	int tx_update_mark;
+	int tx_noupd;
+#endif
+	spinlock_t tx_lock;	/* NETIF_F_LLTX mode */
+
+	/* rarely used */
+	u8 port;
+	u32 msg_enable;
+	int stats_flag;
+	struct bdx_stats hw_stats;
+	struct pci_dev *pdev;
+
+	struct pci_nic *nic;
+
+	u8 txd_size;
+	u8 txf_size;
+	u8 rxd_size;
+	u8 rxf_size;
+	u32 rdintcm;
+	u32 tdintcm;
+};
+
+/* RX FREE descriptor - 64bit*/
+struct rxf_desc {
+	u32 info;		/* Buffer Count + Info - described below */
+	u32 va_lo;		/* VAdr[31:0] */
+	u32 va_hi;		/* VAdr[63:32] */
+	u32 pa_lo;		/* PAdr[31:0] */
+	u32 pa_hi;		/* PAdr[63:32] */
+	u32 len;		/* Buffer Length */
+};
+
+#define GET_RXD_BC(x)			GET_BITS_SHIFT((x), 5, 0)
+#define GET_RXD_RXFQ(x)			GET_BITS_SHIFT((x), 2, 8)
+#define GET_RXD_TO(x)			GET_BITS_SHIFT((x), 1, 15)
+#define GET_RXD_TYPE(x)			GET_BITS_SHIFT((x), 4, 16)
+#define GET_RXD_ERR(x)			GET_BITS_SHIFT((x), 6, 21)
+#define GET_RXD_RXP(x)			GET_BITS_SHIFT((x), 1, 27)
+#define GET_RXD_PKT_ID(x)		GET_BITS_SHIFT((x), 3, 28)
+#define GET_RXD_VTAG(x)			GET_BITS_SHIFT((x), 1, 31)
+#define GET_RXD_VLAN_ID(x)		GET_BITS_SHIFT((x), 12, 0)
+#define GET_RXD_VLAN_TCI(x)		GET_BITS_SHIFT((x), 16, 0)
+#define GET_RXD_CFI(x)			GET_BITS_SHIFT((x), 1, 12)
+#define GET_RXD_PRIO(x)			GET_BITS_SHIFT((x), 3, 13)
+
+struct rxd_desc {
+	u32 rxd_val1;
+	u16 len;
+	u16 rxd_vlan;
+	u32 va_lo;
+	u32 va_hi;
+};
+
+/* PBL describes each virtual buffer to be */
+/* transmitted from the host.*/
+struct pbl {
+	u32 pa_lo;
+	u32 pa_hi;
+	u32 len;
+};
+
+/* First word for TXD descriptor. It means: type = 3 for regular Tx packet,
+ * hw_csum = 7 for ip+udp+tcp hw checksums */
+#define TXD_W1_VAL(bc, checksum, vtag, lgsnd, vlan_id)	\
+	((bc) | ((checksum)<<5) | ((vtag)<<8) | \
+	((lgsnd)<<9) | (0x30000) | ((vlan_id)<<20))
+
+struct txd_desc {
+	u32 txd_val1;
+	u16 mss;
+	u16 length;
+	u32 va_lo;
+	u32 va_hi;
+	struct pbl pbl[];	/* Fragments */
+} __packed;
+
+/* Register region size */
+#define BDX_REGS_SIZE	  0x1000
+
+/* Registers from 0x0000-0x00fc were remapped to 0x4000-0x40fc */
+#define regTXD_CFG1_0   0x4000
+#define regRXF_CFG1_0   0x4010
+#define regRXD_CFG1_0   0x4020
+#define regTXF_CFG1_0   0x4030
+#define regTXD_CFG0_0   0x4040
+#define regRXF_CFG0_0   0x4050
+#define regRXD_CFG0_0   0x4060
+#define regTXF_CFG0_0   0x4070
+#define regTXD_WPTR_0   0x4080
+#define regRXF_WPTR_0   0x4090
+#define regRXD_WPTR_0   0x40A0
+#define regTXF_WPTR_0   0x40B0
+#define regTXD_RPTR_0   0x40C0
+#define regRXF_RPTR_0   0x40D0
+#define regRXD_RPTR_0   0x40E0
+#define regTXF_RPTR_0   0x40F0
+#define regTXF_RPTR_3   0x40FC
+
+/* hardware versioning */
+#define  FW_VER         0x5010
+#define  SROM_VER       0x5020
+#define  FPGA_VER       0x5030
+#define  FPGA_SEED      0x5040
+
+/* Registers from 0x0100-0x0150 were remapped to 0x5100-0x5150 */
+#define regISR regISR0
+#define regISR0          0x5100
+
+#define regIMR regIMR0
+#define regIMR0          0x5110
+
+#define regRDINTCM0      0x5120
+#define regRDINTCM2      0x5128
+
+#define regTDINTCM0      0x5130
+
+#define regISR_MSK0      0x5140
+
+#define regINIT_SEMAPHORE 0x5170
+#define regINIT_STATUS    0x5180
+
+#define regMAC_LNK_STAT  0x0200
+#define MAC_LINK_STAT    0x4	/* Link state */
+
+#define regGMAC_RXF_A   0x1240
+
+#define regUNC_MAC0_A   0x1250
+#define regUNC_MAC1_A   0x1260
+#define regUNC_MAC2_A   0x1270
+
+#define regVLAN_0       0x1800
+
+#define regMAX_FRAME_A  0x12C0
+
+#define regRX_MAC_MCST0    0x1A80
+#define regRX_MAC_MCST1    0x1A84
+#define MAC_MCST_NUM       15
+#define regRX_MCST_HASH0   0x1A00
+#define MAC_MCST_HASH_NUM  8
+
+#define regVPC                  0x2300
+#define regVIC                  0x2320
+#define regVGLB                 0x2340
+
+#define regCLKPLL               0x5000
+
+/*for 10G only*/
+#define regREVISION        0x6000
+#define regSCRATCH         0x6004
+#define regCTRLST          0x6008
+#define regMAC_ADDR_0      0x600C
+#define regMAC_ADDR_1      0x6010
+#define regFRM_LENGTH      0x6014
+#define regPAUSE_QUANT     0x6018
+#define regRX_FIFO_SECTION 0x601C
+#define regTX_FIFO_SECTION 0x6020
+#define regRX_FULLNESS     0x6024
+#define regTX_FULLNESS     0x6028
+#define regHASHTABLE       0x602C
+#define regMDIO_ST         0x6030
+#define regMDIO_CTL        0x6034
+#define regMDIO_DATA       0x6038
+#define regMDIO_ADDR       0x603C
+
+#define regRST_PORT        0x7000
+#define regDIS_PORT        0x7010
+#define regRST_QU          0x7020
+#define regDIS_QU          0x7030
+
+#define regCTRLST_TX_ENA   0x0001
+#define regCTRLST_RX_ENA   0x0002
+#define regCTRLST_PRM_ENA  0x0010
+#define regCTRLST_PAD_ENA  0x0020
+
+#define regCTRLST_BASE     (regCTRLST_PAD_ENA|regCTRLST_PRM_ENA)
+
+#define regRX_FLT   0x1400
+
+/* TXD TXF RXF RXD  CONFIG 0x0000 --- 0x007c*/
+#define  TX_RX_CFG1_BASE          0xffffffff	/*0-31 */
+#define  TX_RX_CFG0_BASE          0xfffff000	/*31:12 */
+#define  TX_RX_CFG0_RSVD          0x0ffc	/*11:2 */
+#define  TX_RX_CFG0_SIZE          0x0003	/*1:0 */
+
+/*  TXD TXF RXF RXD  WRITE 0x0080 --- 0x00BC */
+#define  TXF_WPTR_WR_PTR        0x7ff8	/*14:3 */
+
+/*  TXD TXF RXF RXD  READ  0x00CO --- 0x00FC */
+#define  TXF_RPTR_RD_PTR        0x7ff8	/*14:3 */
+
+#define TXF_WPTR_MASK 0x7ff0	/* last 4 bits are dropped
+				 * size is rounded to 16 */
+
+/*  regISR 0x0100 */
+/*  regIMR 0x0110 */
+#define  IMR_INPROG   0x80000000	/*31 */
+#define  IR_LNKCHG1   0x10000000	/*28 */
+#define  IR_LNKCHG0   0x08000000	/*27 */
+#define  IR_GPIO      0x04000000	/*26 */
+#define  IR_RFRSH     0x02000000	/*25 */
+#define  IR_RSVD      0x01000000	/*24 */
+#define  IR_SWI       0x00800000	/*23 */
+#define  IR_RX_FREE_3 0x00400000	/*22 */
+#define  IR_RX_FREE_2 0x00200000	/*21 */
+#define  IR_RX_FREE_1 0x00100000	/*20 */
+#define  IR_RX_FREE_0 0x00080000	/*19 */
+#define  IR_TX_FREE_3 0x00040000	/*18 */
+#define  IR_TX_FREE_2 0x00020000	/*17 */
+#define  IR_TX_FREE_1 0x00010000	/*16 */
+#define  IR_TX_FREE_0 0x00008000	/*15 */
+#define  IR_RX_DESC_3 0x00004000	/*14 */
+#define  IR_RX_DESC_2 0x00002000	/*13 */
+#define  IR_RX_DESC_1 0x00001000	/*12 */
+#define  IR_RX_DESC_0 0x00000800	/*11 */
+#define  IR_PSE       0x00000400	/*10 */
+#define  IR_TMR3      0x00000200	/*9 */
+#define  IR_TMR2      0x00000100	/*8 */
+#define  IR_TMR1      0x00000080	/*7 */
+#define  IR_TMR0      0x00000040	/*6 */
+#define  IR_VNT       0x00000020	/*5 */
+#define  IR_RxFL      0x00000010	/*4 */
+#define  IR_SDPERR    0x00000008	/*3 */
+#define  IR_TR        0x00000004	/*2 */
+#define  IR_PCIE_LINK 0x00000002	/*1 */
+#define  IR_PCIE_TOUT 0x00000001	/*0 */
+
+#define  IR_EXTRA (IR_RX_FREE_0 | IR_LNKCHG0 | IR_PSE | \
+    IR_TMR0 | IR_PCIE_LINK | IR_PCIE_TOUT)
+#define  IR_RUN (IR_EXTRA | IR_RX_DESC_0 | IR_TX_FREE_0)
+#define  IR_ALL 0xfdfffff7
+
+#define  IR_LNKCHG0_ofst        27
+
+#define  GMAC_RX_FILTER_OSEN  0x1000	/* shared OS enable */
+#define  GMAC_RX_FILTER_TXFC  0x0400	/* Tx flow control */
+#define  GMAC_RX_FILTER_RSV0  0x0200	/* reserved */
+#define  GMAC_RX_FILTER_FDA   0x0100	/* filter out direct address */
+#define  GMAC_RX_FILTER_AOF   0x0080	/* accept over run */
+#define  GMAC_RX_FILTER_ACF   0x0040	/* accept control frames */
+#define  GMAC_RX_FILTER_ARUNT 0x0020	/* accept under run */
+#define  GMAC_RX_FILTER_ACRC  0x0010	/* accept crc error */
+#define  GMAC_RX_FILTER_AM    0x0008	/* accept multicast */
+#define  GMAC_RX_FILTER_AB    0x0004	/* accept broadcast */
+#define  GMAC_RX_FILTER_PRM   0x0001	/* [0:1] promiscuous mode */
+
+#define  MAX_FRAME_AB_VAL       0x3fff	/* 13:0 */
+
+#define  CLKPLL_PLLLKD          0x0200	/*9 */
+#define  CLKPLL_RSTEND          0x0100	/*8 */
+#define  CLKPLL_SFTRST          0x0001	/*0 */
+
+#define  CLKPLL_LKD             (CLKPLL_PLLLKD|CLKPLL_RSTEND)
+
+/*
+ * PCI-E Device Control Register (Offset 0x88)
+ * Source: Luxor Data Sheet, 7.1.3.3.3
+ */
+#define PCI_DEV_CTRL_REG 0x88
+#define GET_DEV_CTRL_MAXPL(x)           GET_BITS_SHIFT(x, 3, 5)
+#define GET_DEV_CTRL_MRRS(x)            GET_BITS_SHIFT(x, 3, 12)
+
+/*
+ * PCI-E Link Status Register (Offset 0x92)
+ * Source: Luxor Data Sheet, 7.1.3.3.7
+ */
+#define PCI_LINK_STATUS_REG 0x92
+#define GET_LINK_STATUS_LANES(x)		GET_BITS_SHIFT(x, 6, 4)
+
+/* Debugging Macros */
+
+#define DBG2(fmt, args...)					\
+	pr_err("%s:%-5d: " fmt, __func__, __LINE__, ## args)
+
+#define BDX_ASSERT(x) BUG_ON(x)
+
+#ifdef DEBUG
+
+#define ENTER						\
+do {							\
+	pr_err("%s:%-5d: ENTER\n", __func__, __LINE__); \
+} while (0)
+
+#define RET(args...)					 \
+do {							 \
+	pr_err("%s:%-5d: RETURN\n", __func__, __LINE__); \
+	return args;					 \
+} while (0)
+
+#define DBG(fmt, args...)					\
+	pr_err("%s:%-5d: " fmt, __func__, __LINE__, ## args)
+#else
+#define ENTER do {  } while (0)
+#define RET(args...)   return args
+#define DBG(fmt, args...)			\
+do {						\
+	if (0)					\
+		pr_err(fmt, ##args);		\
+} while (0)
+#endif
+
+#endif /* _BDX__H */