[vmxnet3] Add VMware vmxnet3 driver

Reviewed-by: Pete Holland <pholland27@gmail.com>
Signed-off-by: Michael Brown <mcb30@ipxe.org>

1 files changed, 669 insertions, 0 deletions

diff --git a/src/drivers/net/vmxnet3.c b/src/drivers/net/vmxnet3.c
new file mode 100644
index 00000000..79cf1d80
--- /dev/null
+++ b/src/drivers/net/vmxnet3.c
@@ -0,0 +1,669 @@
+/*
+ * Copyright (C) 2011 Michael Brown <mbrown@fensystems.co.uk>.
+ *
+ * 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 any later version.
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
+ */
+
+FILE_LICENCE ( GPL2_OR_LATER );
+
+#include <stdint.h>
+#include <errno.h>
+#include <assert.h>
+#include <byteswap.h>
+#include <ipxe/pci.h>
+#include <ipxe/io.h>
+#include <ipxe/malloc.h>
+#include <ipxe/iobuf.h>
+#include <ipxe/netdevice.h>
+#include <ipxe/if_ether.h>
+#include <ipxe/ethernet.h>
+#include "vmxnet3.h"
+
+/**
+ * @file
+ *
+ * VMware vmxnet3 virtual NIC driver
+ *
+ */
+
+/**
+ * Issue command
+ *
+ * @v vmxnet		vmxnet3 NIC
+ * @v command		Command to issue
+ * @ret result		Command result
+ */
+static inline uint32_t vmxnet3_command ( struct vmxnet3_nic *vmxnet,
+					 uint32_t command ) {
+
+	/* Issue command */
+	writel ( command, ( vmxnet->vd + VMXNET3_VD_CMD ) );
+	return readl ( vmxnet->vd + VMXNET3_VD_CMD );
+}
+
+/**
+ * Transmit packet
+ *
+ * @v netdev		Network device
+ * @v iobuf		I/O buffer
+ * @ret rc		Return status code
+ */
+static int vmxnet3_transmit ( struct net_device *netdev,
+			      struct io_buffer *iobuf ) {
+	struct vmxnet3_nic *vmxnet = netdev_priv ( netdev );
+	struct vmxnet3_tx_desc *tx_desc;
+	unsigned int desc_idx;
+	unsigned int generation;
+
+	/* Check that we have a free transmit descriptor */
+	desc_idx = ( vmxnet->count.tx_prod % VMXNET3_NUM_TX_DESC );
+	generation = ( ( vmxnet->count.tx_prod & VMXNET3_NUM_TX_DESC ) ?
+		       0 : cpu_to_le32 ( VMXNET3_TXF_GEN ) );
+	if ( vmxnet->tx_iobuf[desc_idx] ) {
+		DBGC ( vmxnet, "VMXNET3 %p out of transmit descriptors\n",
+		       vmxnet );
+		return -ENOBUFS;
+	}
+
+	/* Increment producer counter */
+	vmxnet->count.tx_prod++;
+
+	/* Store I/O buffer for later completion */
+	vmxnet->tx_iobuf[desc_idx] = iobuf;
+
+	/* Populate transmit descriptor */
+	tx_desc = &vmxnet->dma->tx_desc[desc_idx];
+	tx_desc->address = cpu_to_le64 ( virt_to_bus ( iobuf->data ) );
+	tx_desc->flags[0] = ( generation | cpu_to_le32 ( iob_len ( iobuf ) ) );
+	tx_desc->flags[1] = cpu_to_le32 ( VMXNET3_TXF_CQ | VMXNET3_TXF_EOP );
+
+	/* Hand over descriptor to NIC */
+	wmb();
+	writel ( ( vmxnet->count.tx_prod % VMXNET3_NUM_TX_DESC ),
+		 ( vmxnet->pt + VMXNET3_PT_TXPROD ) );
+
+	return 0;
+}
+
+/**
+ * Poll for completed transmissions
+ *
+ * @v netdev		Network device
+ */
+static void vmxnet3_poll_tx ( struct net_device *netdev ) {
+	struct vmxnet3_nic *vmxnet = netdev_priv ( netdev );
+	struct vmxnet3_tx_comp *tx_comp;
+	struct io_buffer *iobuf;
+	unsigned int comp_idx;
+	unsigned int desc_idx;
+	unsigned int generation;
+
+	while ( 1 ) {
+
+		/* Look for completed descriptors */
+		comp_idx = ( vmxnet->count.tx_cons % VMXNET3_NUM_TX_COMP );
+		generation = ( ( vmxnet->count.tx_cons & VMXNET3_NUM_TX_COMP ) ?
+			       0 : cpu_to_le32 ( VMXNET3_TXCF_GEN ) );
+		tx_comp = &vmxnet->dma->tx_comp[comp_idx];
+		if ( generation != ( tx_comp->flags &
+				     cpu_to_le32 ( VMXNET3_TXCF_GEN ) ) ) {
+			break;
+		}
+
+		/* Increment consumer counter */
+		vmxnet->count.tx_cons++;
+
+		/* Locate corresponding transmit descriptor */
+		desc_idx = ( le32_to_cpu ( tx_comp->index ) %
+			     VMXNET3_NUM_TX_DESC );
+		iobuf = vmxnet->tx_iobuf[desc_idx];
+		if ( ! iobuf ) {
+			DBGC ( vmxnet, "VMXNET3 %p completed on empty transmit "
+			       "buffer %#x/%#x\n", vmxnet, comp_idx, desc_idx );
+			netdev_tx_err ( netdev, NULL, -ENOTTY );
+			continue;
+		}
+
+		/* Remove I/O buffer from transmit queue */
+		vmxnet->tx_iobuf[desc_idx] = NULL;
+
+		/* Report transmission completion to network layer */
+		DBGC2 ( vmxnet, "VMXNET3 %p completed TX %#x/%#x (len %#zx)\n",
+			vmxnet, comp_idx, desc_idx, iob_len ( iobuf ) );
+		netdev_tx_complete ( netdev, iobuf );
+	}
+}
+
+/**
+ * Flush any uncompleted transmit buffers
+ *
+ * @v netdev		Network device
+ */
+static void vmxnet3_flush_tx ( struct net_device *netdev ) {
+	struct vmxnet3_nic *vmxnet = netdev_priv ( netdev );
+	unsigned int i;
+
+	for ( i = 0 ; i < VMXNET3_NUM_TX_DESC ; i++ ) {
+		if ( vmxnet->tx_iobuf[i] ) {
+			netdev_tx_complete_err ( netdev, vmxnet->tx_iobuf[i],
+						 -ECANCELED );
+			vmxnet->tx_iobuf[i] = NULL;
+		}
+	}
+}
+
+/**
+ * Refill receive ring
+ *
+ * @v netdev		Network device
+ */
+static void vmxnet3_refill_rx ( struct net_device *netdev ) {
+	struct vmxnet3_nic *vmxnet = netdev_priv ( netdev );
+	struct vmxnet3_rx_desc *rx_desc;
+	struct io_buffer *iobuf;
+	unsigned int orig_rx_prod = vmxnet->count.rx_prod;
+	unsigned int desc_idx;
+	unsigned int generation;
+
+	/* Fill receive ring to specified fill level */
+	while ( vmxnet->count.rx_fill < VMXNET3_RX_FILL ) {
+
+		/* Locate receive descriptor */
+		desc_idx = ( vmxnet->count.rx_prod % VMXNET3_NUM_RX_DESC );
+		generation = ( ( vmxnet->count.rx_prod & VMXNET3_NUM_RX_DESC ) ?
+			       0 : cpu_to_le32 ( VMXNET3_RXF_GEN ) );
+		assert ( vmxnet->rx_iobuf[desc_idx] == NULL );
+
+		/* Allocate I/O buffer */
+		iobuf = alloc_iob ( VMXNET3_MTU + NET_IP_ALIGN );
+		if ( ! iobuf ) {
+			/* Non-fatal low memory condition */
+			break;
+		}
+		iob_reserve ( iobuf, NET_IP_ALIGN );
+
+		/* Increment producer counter and fill level */
+		vmxnet->count.rx_prod++;
+		vmxnet->count.rx_fill++;
+
+		/* Store I/O buffer for later completion */
+		vmxnet->rx_iobuf[desc_idx] = iobuf;
+
+		/* Populate receive descriptor */
+		rx_desc = &vmxnet->dma->rx_desc[desc_idx];
+		rx_desc->address = cpu_to_le64 ( virt_to_bus ( iobuf->data ) );
+		rx_desc->flags = ( generation | cpu_to_le32 ( VMXNET3_MTU ) );
+
+	}
+
+	/* Hand over any new descriptors to NIC */
+	if ( vmxnet->count.rx_prod != orig_rx_prod ) {
+		wmb();
+		writel ( ( vmxnet->count.rx_prod % VMXNET3_NUM_RX_DESC ),
+			 ( vmxnet->pt + VMXNET3_PT_RXPROD ) );
+	}
+}
+
+/**
+ * Poll for received packets
+ *
+ * @v netdev		Network device
+ */
+static void vmxnet3_poll_rx ( struct net_device *netdev ) {
+	struct vmxnet3_nic *vmxnet = netdev_priv ( netdev );
+	struct vmxnet3_rx_comp *rx_comp;
+	struct io_buffer *iobuf;
+	unsigned int comp_idx;
+	unsigned int desc_idx;
+	unsigned int generation;
+	size_t len;
+
+	while ( 1 ) {
+
+		/* Look for completed descriptors */
+		comp_idx = ( vmxnet->count.rx_cons % VMXNET3_NUM_RX_COMP );
+		generation = ( ( vmxnet->count.rx_cons & VMXNET3_NUM_RX_COMP ) ?
+			       0 : cpu_to_le32 ( VMXNET3_RXCF_GEN ) );
+		rx_comp = &vmxnet->dma->rx_comp[comp_idx];
+		if ( generation != ( rx_comp->flags &
+				     cpu_to_le32 ( VMXNET3_RXCF_GEN ) ) ) {
+			break;
+		}
+
+		/* Increment consumer counter */
+		vmxnet->count.rx_cons++;
+
+		/* Locate corresponding receive descriptor */
+		desc_idx = ( le32_to_cpu ( rx_comp->index ) %
+			     VMXNET3_NUM_RX_DESC );
+		iobuf = vmxnet->rx_iobuf[desc_idx];
+		if ( ! iobuf ) {
+			DBGC ( vmxnet, "VMXNET3 %p completed on empty receive "
+			       "buffer %#x/%#x\n", vmxnet, comp_idx, desc_idx );
+			netdev_rx_err ( netdev, NULL, -ENOTTY );
+			continue;
+		}
+
+		/* Remove I/O buffer from receive queue */
+		vmxnet->rx_iobuf[desc_idx] = NULL;
+		vmxnet->count.rx_fill--;
+
+		/* Deliver packet to network layer */
+		len = ( le32_to_cpu ( rx_comp->len ) &
+			( VMXNET3_MAX_PACKET_LEN - 1 ) );
+		DBGC2 ( vmxnet, "VMXNET3 %p completed RX %#x/%#x (len %#zx)\n",
+			vmxnet, comp_idx, desc_idx, len );
+		iob_put ( iobuf, len );
+		netdev_rx ( netdev, iobuf );
+	}
+}
+
+/**
+ * Flush any uncompleted receive buffers
+ *
+ * @v netdev		Network device
+ */
+static void vmxnet3_flush_rx ( struct net_device *netdev ) {
+	struct vmxnet3_nic *vmxnet = netdev_priv ( netdev );
+	struct io_buffer *iobuf;
+	unsigned int i;
+
+	for ( i = 0 ; i < VMXNET3_NUM_RX_DESC ; i++ ) {
+		if ( ( iobuf = vmxnet->rx_iobuf[i] ) != NULL ) {
+			netdev_rx_err ( netdev, iobuf, -ECANCELED );
+			vmxnet->rx_iobuf[i] = NULL;
+		}
+	}
+}
+
+/**
+ * Check link state
+ *
+ * @v netdev		Network device
+ */
+static void vmxnet3_check_link ( struct net_device *netdev ) {
+	struct vmxnet3_nic *vmxnet = netdev_priv ( netdev );
+	uint32_t state;
+	int link_up;
+	unsigned int link_speed;
+
+	/* Get link state */
+	state = vmxnet3_command ( vmxnet, VMXNET3_CMD_GET_LINK );
+	link_up = ( state & 1 );
+	link_speed = ( state >> 16 );
+
+	/* Report link state to network device */
+	if ( link_up ) {
+		DBGC ( vmxnet, "VMXNET3 %p link is up at %d Mbps\n",
+		       vmxnet, link_speed );
+		netdev_link_up ( netdev );
+	} else {
+		DBGC ( vmxnet, "VMXNET3 %p link is down\n", vmxnet );
+		netdev_link_down ( netdev );
+	}
+}
+
+/**
+ * Poll for events
+ *
+ * @v netdev		Network device
+ */
+static void vmxnet3_poll_events ( struct net_device *netdev ) {
+	struct vmxnet3_nic *vmxnet = netdev_priv ( netdev );
+	uint32_t events;
+
+	/* Do nothing unless there are events to process */
+	if ( ! vmxnet->dma->shared.ecr )
+		return;
+	events = le32_to_cpu ( vmxnet->dma->shared.ecr );
+
+	/* Acknowledge these events */
+	writel ( events, ( vmxnet->vd + VMXNET3_VD_ECR ) );
+
+	/* Check for link state change */
+	if ( events & VMXNET3_ECR_LINK ) {
+		vmxnet3_check_link ( netdev );
+		events &= ~VMXNET3_ECR_LINK;
+	}
+
+	/* Check for queue errors */
+	if ( events & ( VMXNET3_ECR_TQERR | VMXNET3_ECR_RQERR ) ) {
+		vmxnet3_command ( vmxnet, VMXNET3_CMD_GET_QUEUE_STATUS );
+		DBGC ( vmxnet, "VMXNET3 %p queue error status (TX %08x, RX "
+		       "%08x)\n", vmxnet,
+		       le32_to_cpu ( vmxnet->dma->queues.tx.status.error ),
+		       le32_to_cpu ( vmxnet->dma->queues.rx.status.error ) );
+		/* Report errors to allow for visibility via "ifstat" */
+		if ( events & VMXNET3_ECR_TQERR )
+			netdev_tx_err ( netdev, NULL, -EPIPE );
+		if ( events & VMXNET3_ECR_RQERR )
+			netdev_rx_err ( netdev, NULL, -EPIPE );
+		events &= ~( VMXNET3_ECR_TQERR | VMXNET3_ECR_RQERR );
+	}
+
+	/* Check for unknown events */
+	if ( events ) {
+		DBGC ( vmxnet, "VMXNET3 %p unknown events %08x\n",
+		       vmxnet, events );
+		/* Report error to allow for visibility via "ifstat" */
+		netdev_rx_err ( netdev, NULL, -ENODEV );
+	}
+}
+
+/**
+ * Poll network device
+ *
+ * @v netdev		Network device
+ */
+static void vmxnet3_poll ( struct net_device *netdev ) {
+
+	vmxnet3_poll_events ( netdev );
+	vmxnet3_poll_tx ( netdev );
+	vmxnet3_poll_rx ( netdev );
+	vmxnet3_refill_rx ( netdev );
+}
+
+/**
+ * Enable/disable interrupts
+ *
+ * @v netdev		Network device
+ * @v enable		Interrupts should be enabled
+ */
+static void vmxnet3_irq ( struct net_device *netdev, int enable ) {
+	struct vmxnet3_nic *vmxnet = netdev_priv ( netdev );
+
+	DBGC ( vmxnet, "VMXNET3 %p %s IRQ not implemented\n",
+	       vmxnet, ( enable ? "enable" : "disable" ) );
+}
+
+/**
+ * Set MAC address
+ *
+ * @v vmxnet		vmxnet3 NIC
+ * @v ll_addr		Link-layer address to set
+ */
+static void vmxnet3_set_ll_addr ( struct vmxnet3_nic *vmxnet,
+				  const void *ll_addr ) {
+	struct {
+		uint32_t low;
+		uint32_t high;
+	} __attribute__ (( packed )) mac;
+
+	memset ( &mac, 0, sizeof ( mac ) );
+	memcpy ( &mac, ll_addr, ETH_ALEN );
+	writel ( cpu_to_le32 ( mac.low ), ( vmxnet->vd + VMXNET3_VD_MACL ) );
+	writel ( cpu_to_le32 ( mac.high ), ( vmxnet->vd + VMXNET3_VD_MACH ) );
+}
+
+/**
+ * Open NIC
+ *
+ * @v netdev		Network device
+ * @ret rc		Return status code
+ */
+static int vmxnet3_open ( struct net_device *netdev ) {
+	struct vmxnet3_nic *vmxnet = netdev_priv ( netdev );
+	struct vmxnet3_shared *shared;
+	struct vmxnet3_queues *queues;
+	uint64_t shared_bus;
+	uint64_t queues_bus;
+	uint32_t status;
+	int rc;
+
+	/* Allocate DMA areas */
+	vmxnet->dma = malloc_dma ( sizeof ( *vmxnet->dma ), VMXNET3_DMA_ALIGN );
+	if ( ! vmxnet->dma ) {
+		DBGC ( vmxnet, "VMXNET3 %p could not allocate DMA area\n",
+		       vmxnet );
+		rc = -ENOMEM;
+		goto err_alloc_dma;
+	}
+	memset ( vmxnet->dma, 0, sizeof ( *vmxnet->dma ) );
+
+	/* Populate queue descriptors */
+	queues = &vmxnet->dma->queues;
+	queues->tx.cfg.desc_address =
+		cpu_to_le64 ( virt_to_bus ( &vmxnet->dma->tx_desc ) );
+	queues->tx.cfg.comp_address =
+		cpu_to_le64 ( virt_to_bus ( &vmxnet->dma->tx_comp ) );
+	queues->tx.cfg.num_desc = cpu_to_le32 ( VMXNET3_NUM_TX_DESC );
+	queues->tx.cfg.num_comp = cpu_to_le32 ( VMXNET3_NUM_TX_COMP );
+	queues->rx.cfg.desc_address[0] =
+		cpu_to_le64 ( virt_to_bus ( &vmxnet->dma->rx_desc ) );
+	queues->rx.cfg.comp_address =
+		cpu_to_le64 ( virt_to_bus ( &vmxnet->dma->rx_comp ) );
+	queues->rx.cfg.num_desc[0] = cpu_to_le32 ( VMXNET3_NUM_RX_DESC );
+	queues->rx.cfg.num_comp = cpu_to_le32 ( VMXNET3_NUM_RX_COMP );
+	queues_bus = virt_to_bus ( queues );
+	DBGC ( vmxnet, "VMXNET3 %p queue descriptors at %08llx+%zx\n",
+	       vmxnet, queues_bus, sizeof ( *queues ) );
+
+	/* Populate shared area */
+	shared = &vmxnet->dma->shared;
+	shared->magic = cpu_to_le32 ( VMXNET3_SHARED_MAGIC );
+	shared->misc.version = cpu_to_le32 ( VMXNET3_VERSION_MAGIC );
+	shared->misc.version_support = cpu_to_le32 ( VMXNET3_VERSION_SELECT );
+	shared->misc.upt_version_support =
+		cpu_to_le32 ( VMXNET3_UPT_VERSION_SELECT );
+	shared->misc.queue_desc_address = cpu_to_le64 ( queues_bus );
+	shared->misc.queue_desc_len = cpu_to_le32 ( sizeof ( *queues ) );
+	shared->misc.mtu = cpu_to_le32 ( VMXNET3_MTU );
+	shared->misc.num_tx_queues = 1;
+	shared->misc.num_rx_queues = 1;
+	shared->interrupt.num_intrs = 1;
+	shared->interrupt.control = cpu_to_le32 ( VMXNET3_IC_DISABLE_ALL );
+	shared->rx_filter.mode = cpu_to_le32 ( VMXNET3_RXM_UCAST |
+					       VMXNET3_RXM_BCAST |
+					       VMXNET3_RXM_ALL_MULTI );
+	shared_bus = virt_to_bus ( shared );
+	DBGC ( vmxnet, "VMXNET3 %p shared area at %08llx+%zx\n",
+	       vmxnet, shared_bus, sizeof ( *shared ) );
+
+	/* Zero counters */
+	memset ( &vmxnet->count, 0, sizeof ( vmxnet->count ) );
+
+	/* Set MAC address */
+	vmxnet3_set_ll_addr ( vmxnet, &netdev->ll_addr );
+
+	/* Pass shared area to device */
+	writel ( ( shared_bus >> 0 ), ( vmxnet->vd + VMXNET3_VD_DSAL ) );
+	writel ( ( shared_bus >> 32 ), ( vmxnet->vd + VMXNET3_VD_DSAH ) );
+
+	/* Activate device */
+	if ( ( status = vmxnet3_command ( vmxnet,
+					  VMXNET3_CMD_ACTIVATE_DEV ) ) != 0 ) {
+		DBGC ( vmxnet, "VMXNET3 %p could not activate (status %#x)\n",
+		       vmxnet, status );
+		rc = -EIO;
+		goto err_activate;
+	}
+
+	/* Fill receive ring */
+	vmxnet3_refill_rx ( netdev );
+
+	return 0;
+
+	vmxnet3_command ( vmxnet, VMXNET3_CMD_QUIESCE_DEV );
+	vmxnet3_command ( vmxnet, VMXNET3_CMD_RESET_DEV );
+ err_activate:
+	vmxnet3_flush_tx ( netdev );
+	vmxnet3_flush_rx ( netdev );
+	free_dma ( vmxnet->dma, sizeof ( *vmxnet->dma ) );
+ err_alloc_dma:
+	return rc;
+}
+
+/**
+ * Close NIC
+ *
+ * @v netdev		Network device
+ */
+static void vmxnet3_close ( struct net_device *netdev ) {
+	struct vmxnet3_nic *vmxnet = netdev_priv ( netdev );
+
+	vmxnet3_command ( vmxnet, VMXNET3_CMD_QUIESCE_DEV );
+	vmxnet3_command ( vmxnet, VMXNET3_CMD_RESET_DEV );
+	vmxnet3_flush_tx ( netdev );
+	vmxnet3_flush_rx ( netdev );
+	free_dma ( vmxnet->dma, sizeof ( *vmxnet->dma ) );
+}
+
+/** vmxnet3 net device operations */
+static struct net_device_operations vmxnet3_operations = {
+	.open		= vmxnet3_open,
+	.close		= vmxnet3_close,
+	.transmit	= vmxnet3_transmit,
+	.poll		= vmxnet3_poll,
+	.irq		= vmxnet3_irq,
+};
+
+/**
+ * Check version
+ *
+ * @v vmxnet		vmxnet3 NIC
+ * @ret rc		Return status code
+ */
+static int vmxnet3_check_version ( struct vmxnet3_nic *vmxnet ) {
+	uint32_t version;
+	uint32_t upt_version;
+
+	/* Read version */
+	version = readl ( vmxnet->vd + VMXNET3_VD_VRRS );
+	upt_version = readl ( vmxnet->vd + VMXNET3_VD_UVRS );
+	DBGC ( vmxnet, "VMXNET3 %p is version %d (UPT version %d)\n",
+	       vmxnet, version, upt_version );
+
+	/* Inform NIC of driver version */
+	writel ( VMXNET3_VERSION_SELECT, ( vmxnet->vd + VMXNET3_VD_VRRS ) );
+	writel ( VMXNET3_UPT_VERSION_SELECT, ( vmxnet->vd + VMXNET3_VD_UVRS ) );
+
+	return 0;
+}
+
+/**
+ * Get permanent MAC address
+ *
+ * @v vmxnet		vmxnet3 NIC
+ * @v hw_addr		Hardware address to fill in
+ */
+static void vmxnet3_get_hw_addr ( struct vmxnet3_nic *vmxnet, void *hw_addr ) {
+	struct {
+		uint32_t low;
+		uint32_t high;
+	} __attribute__ (( packed )) mac;
+
+	mac.low = le32_to_cpu ( vmxnet3_command ( vmxnet,
+					       VMXNET3_CMD_GET_PERM_MAC_LO ) );
+	mac.high = le32_to_cpu ( vmxnet3_command ( vmxnet,
+					       VMXNET3_CMD_GET_PERM_MAC_HI ) );
+	memcpy ( hw_addr, &mac, ETH_ALEN );
+}
+
+/**
+ * Probe PCI device
+ *
+ * @v pci		PCI device
+ * @v id		PCI ID
+ * @ret rc		Return status code
+ */
+static int vmxnet3_probe ( struct pci_device *pci ) {
+	struct net_device *netdev;
+	struct vmxnet3_nic *vmxnet;
+	int rc;
+
+	/* Allocate network device */
+	netdev = alloc_etherdev ( sizeof ( *vmxnet ) );
+	if ( ! netdev ) {
+		rc = -ENOMEM;
+		goto err_alloc_etherdev;
+	}
+	netdev_init ( netdev, &vmxnet3_operations );
+	vmxnet = netdev_priv ( netdev );
+	pci_set_drvdata ( pci, netdev );
+	netdev->dev = &pci->dev;
+	memset ( vmxnet, 0, sizeof ( *vmxnet ) );
+
+	/* Fix up PCI device */
+	adjust_pci_device ( pci );
+
+	/* Map PCI BARs */
+	vmxnet->pt = ioremap ( pci_bar_start ( pci, VMXNET3_PT_BAR ),
+			       VMXNET3_PT_LEN );
+	vmxnet->vd = ioremap ( pci_bar_start ( pci, VMXNET3_VD_BAR ),
+			       VMXNET3_VD_LEN );
+
+	/* Version check */
+	if ( ( rc = vmxnet3_check_version ( vmxnet ) ) != 0 )
+		goto err_check_version;
+
+	/* Reset device */
+	if ( ( rc = vmxnet3_command ( vmxnet, VMXNET3_CMD_RESET_DEV ) ) != 0 )
+		goto err_reset;
+
+	/* Read initial MAC address */
+	vmxnet3_get_hw_addr ( vmxnet, &netdev->hw_addr );
+
+	/* Register network device */
+	if ( ( rc = register_netdev ( netdev ) ) != 0 ) {
+		DBGC ( vmxnet, "VMXNET3 %p could not register net device: "
+		       "%s\n", vmxnet, strerror ( rc ) );
+		goto err_register_netdev;
+	}
+
+	/* Get initial link state */
+	vmxnet3_check_link ( netdev );
+
+	return 0;
+
+	unregister_netdev ( netdev );
+ err_register_netdev:
+ err_reset:
+ err_check_version:
+	iounmap ( vmxnet->vd );
+	iounmap ( vmxnet->pt );
+	netdev_nullify ( netdev );
+	netdev_put ( netdev );
+ err_alloc_etherdev:
+	return rc;
+}
+
+/**
+ * Remove PCI device
+ *
+ * @v pci		PCI device
+ */
+static void vmxnet3_remove ( struct pci_device *pci ) {
+	struct net_device *netdev = pci_get_drvdata ( pci );
+	struct vmxnet3_nic *vmxnet = netdev_priv ( netdev );
+
+	unregister_netdev ( netdev );
+	iounmap ( vmxnet->vd );
+	iounmap ( vmxnet->pt );
+	netdev_nullify ( netdev );
+	netdev_put ( netdev );
+}
+
+/** vmxnet3 PCI IDs */
+static struct pci_device_id vmxnet3_nics[] = {
+	PCI_ROM ( 0x15ad, 0x07b0, "vmxnet3", "vmxnet3 virtual NIC", 0 ),
+};
+
+/** vmxnet3 PCI driver */
+struct pci_driver vmxnet3_driver __pci_driver = {
+	.ids = vmxnet3_nics,
+	.id_count = ( sizeof ( vmxnet3_nics ) / sizeof ( vmxnet3_nics[0] ) ),
+	.probe = vmxnet3_probe,
+	.remove = vmxnet3_remove,
+};