net: VMware virtual Ethernet NIC driver: vmxnet3

Ethernet NIC driver for VMware's vmxnet3

From: Shreyas Bhatewara <sbhatewara@vmware.com>

This patch adds driver support for VMware's virtual Ethernet NIC: vmxnet3
Guests running on VMware hypervisors supporting vmxnet3 device will thus have
access to improved network functionalities and performance.

Signed-off-by: Shreyas Bhatewara <sbhatewara@vmware.com>
Signed-off-by: Bhavesh Davda <bhavesh@vmware.com>
Signed-off-by: Ronghua Zhang <ronghua@vmware.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 2556 insertions, 0 deletions

diff --git a/drivers/net/vmxnet3/vmxnet3_drv.c b/drivers/net/vmxnet3/vmxnet3_drv.c
new file mode 100644
index 00000000000..44fb0c5a280
--- /dev/null
+++ b/drivers/net/vmxnet3/vmxnet3_drv.c
@@ -0,0 +1,2556 @@
+/*
+ * Linux driver for VMware's vmxnet3 ethernet NIC.
+ *
+ * Copyright (C) 2008-2009, VMware, Inc. All Rights Reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the
+ * Free Software Foundation; version 2 of the License and no 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, GOOD TITLE or
+ * NON INFRINGEMENT. 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., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
+ *
+ * The full GNU General Public License is included in this distribution in
+ * the file called "COPYING".
+ *
+ * Maintained by: Shreyas Bhatewara <pv-drivers@vmware.com>
+ *
+ */
+
+#include "vmxnet3_int.h"
+
+char vmxnet3_driver_name[] = "vmxnet3";
+#define VMXNET3_DRIVER_DESC "VMware vmxnet3 virtual NIC driver"
+
+
+/*
+ * PCI Device ID Table
+ * Last entry must be all 0s
+ */
+static const struct pci_device_id vmxnet3_pciid_table[] = {
+	{PCI_VDEVICE(VMWARE, PCI_DEVICE_ID_VMWARE_VMXNET3)},
+	{0}
+};
+
+MODULE_DEVICE_TABLE(pci, vmxnet3_pciid_table);
+
+static atomic_t devices_found;
+
+
+/*
+ *    Enable/Disable the given intr
+ */
+static void
+vmxnet3_enable_intr(struct vmxnet3_adapter *adapter, unsigned intr_idx)
+{
+	VMXNET3_WRITE_BAR0_REG(adapter, VMXNET3_REG_IMR + intr_idx * 8, 0);
+}
+
+
+static void
+vmxnet3_disable_intr(struct vmxnet3_adapter *adapter, unsigned intr_idx)
+{
+	VMXNET3_WRITE_BAR0_REG(adapter, VMXNET3_REG_IMR + intr_idx * 8, 1);
+}
+
+
+/*
+ *    Enable/Disable all intrs used by the device
+ */
+static void
+vmxnet3_enable_all_intrs(struct vmxnet3_adapter *adapter)
+{
+	int i;
+
+	for (i = 0; i < adapter->intr.num_intrs; i++)
+		vmxnet3_enable_intr(adapter, i);
+}
+
+
+static void
+vmxnet3_disable_all_intrs(struct vmxnet3_adapter *adapter)
+{
+	int i;
+
+	for (i = 0; i < adapter->intr.num_intrs; i++)
+		vmxnet3_disable_intr(adapter, i);
+}
+
+
+static void
+vmxnet3_ack_events(struct vmxnet3_adapter *adapter, u32 events)
+{
+	VMXNET3_WRITE_BAR1_REG(adapter, VMXNET3_REG_ECR, events);
+}
+
+
+static bool
+vmxnet3_tq_stopped(struct vmxnet3_tx_queue *tq, struct vmxnet3_adapter *adapter)
+{
+	return netif_queue_stopped(adapter->netdev);
+}
+
+
+static void
+vmxnet3_tq_start(struct vmxnet3_tx_queue *tq, struct vmxnet3_adapter *adapter)
+{
+	tq->stopped = false;
+	netif_start_queue(adapter->netdev);
+}
+
+
+static void
+vmxnet3_tq_wake(struct vmxnet3_tx_queue *tq, struct vmxnet3_adapter *adapter)
+{
+	tq->stopped = false;
+	netif_wake_queue(adapter->netdev);
+}
+
+
+static void
+vmxnet3_tq_stop(struct vmxnet3_tx_queue *tq, struct vmxnet3_adapter *adapter)
+{
+	tq->stopped = true;
+	tq->num_stop++;
+	netif_stop_queue(adapter->netdev);
+}
+
+
+/*
+ * Check the link state. This may start or stop the tx queue.
+ */
+static void
+vmxnet3_check_link(struct vmxnet3_adapter *adapter)
+{
+	u32 ret;
+
+	VMXNET3_WRITE_BAR1_REG(adapter, VMXNET3_REG_CMD, VMXNET3_CMD_GET_LINK);
+	ret = VMXNET3_READ_BAR1_REG(adapter, VMXNET3_REG_CMD);
+	adapter->link_speed = ret >> 16;
+	if (ret & 1) { /* Link is up. */
+		printk(KERN_INFO "%s: NIC Link is Up %d Mbps\n",
+		       adapter->netdev->name, adapter->link_speed);
+		if (!netif_carrier_ok(adapter->netdev))
+			netif_carrier_on(adapter->netdev);
+
+		vmxnet3_tq_start(&adapter->tx_queue, adapter);
+	} else {
+		printk(KERN_INFO "%s: NIC Link is Down\n",
+		       adapter->netdev->name);
+		if (netif_carrier_ok(adapter->netdev))
+			netif_carrier_off(adapter->netdev);
+
+		vmxnet3_tq_stop(&adapter->tx_queue, adapter);
+	}
+}
+
+
+static void
+vmxnet3_process_events(struct vmxnet3_adapter *adapter)
+{
+	u32 events = adapter->shared->ecr;
+	if (!events)
+		return;
+
+	vmxnet3_ack_events(adapter, events);
+
+	/* Check if link state has changed */
+	if (events & VMXNET3_ECR_LINK)
+		vmxnet3_check_link(adapter);
+
+	/* Check if there is an error on xmit/recv queues */
+	if (events & (VMXNET3_ECR_TQERR | VMXNET3_ECR_RQERR)) {
+		VMXNET3_WRITE_BAR1_REG(adapter, VMXNET3_REG_CMD,
+				       VMXNET3_CMD_GET_QUEUE_STATUS);
+
+		if (adapter->tqd_start->status.stopped) {
+			printk(KERN_ERR "%s: tq error 0x%x\n",
+			       adapter->netdev->name,
+			       adapter->tqd_start->status.error);
+		}
+		if (adapter->rqd_start->status.stopped) {
+			printk(KERN_ERR "%s: rq error 0x%x\n",
+			       adapter->netdev->name,
+			       adapter->rqd_start->status.error);
+		}
+
+		schedule_work(&adapter->work);
+	}
+}
+
+
+static void
+vmxnet3_unmap_tx_buf(struct vmxnet3_tx_buf_info *tbi,
+		     struct pci_dev *pdev)
+{
+	if (tbi->map_type == VMXNET3_MAP_SINGLE)
+		pci_unmap_single(pdev, tbi->dma_addr, tbi->len,
+				 PCI_DMA_TODEVICE);
+	else if (tbi->map_type == VMXNET3_MAP_PAGE)
+		pci_unmap_page(pdev, tbi->dma_addr, tbi->len,
+			       PCI_DMA_TODEVICE);
+	else
+		BUG_ON(tbi->map_type != VMXNET3_MAP_NONE);
+
+	tbi->map_type = VMXNET3_MAP_NONE; /* to help debugging */
+}
+
+
+static int
+vmxnet3_unmap_pkt(u32 eop_idx, struct vmxnet3_tx_queue *tq,
+		  struct pci_dev *pdev,	struct vmxnet3_adapter *adapter)
+{
+	struct sk_buff *skb;
+	int entries = 0;
+
+	/* no out of order completion */
+	BUG_ON(tq->buf_info[eop_idx].sop_idx != tq->tx_ring.next2comp);
+	BUG_ON(tq->tx_ring.base[eop_idx].txd.eop != 1);
+
+	skb = tq->buf_info[eop_idx].skb;
+	BUG_ON(skb == NULL);
+	tq->buf_info[eop_idx].skb = NULL;
+
+	VMXNET3_INC_RING_IDX_ONLY(eop_idx, tq->tx_ring.size);
+
+	while (tq->tx_ring.next2comp != eop_idx) {
+		vmxnet3_unmap_tx_buf(tq->buf_info + tq->tx_ring.next2comp,
+				     pdev);
+
+		/* update next2comp w/o tx_lock. Since we are marking more,
+		 * instead of less, tx ring entries avail, the worst case is
+		 * that the tx routine incorrectly re-queues a pkt due to
+		 * insufficient tx ring entries.
+		 */
+		vmxnet3_cmd_ring_adv_next2comp(&tq->tx_ring);
+		entries++;
+	}
+
+	dev_kfree_skb_any(skb);
+	return entries;
+}
+
+
+static int
+vmxnet3_tq_tx_complete(struct vmxnet3_tx_queue *tq,
+			struct vmxnet3_adapter *adapter)
+{
+	int completed = 0;
+	union Vmxnet3_GenericDesc *gdesc;
+
+	gdesc = tq->comp_ring.base + tq->comp_ring.next2proc;
+	while (gdesc->tcd.gen == tq->comp_ring.gen) {
+		completed += vmxnet3_unmap_pkt(gdesc->tcd.txdIdx, tq,
+					       adapter->pdev, adapter);
+
+		vmxnet3_comp_ring_adv_next2proc(&tq->comp_ring);
+		gdesc = tq->comp_ring.base + tq->comp_ring.next2proc;
+	}
+
+	if (completed) {
+		spin_lock(&tq->tx_lock);
+		if (unlikely(vmxnet3_tq_stopped(tq, adapter) &&
+			     vmxnet3_cmd_ring_desc_avail(&tq->tx_ring) >
+			     VMXNET3_WAKE_QUEUE_THRESHOLD(tq) &&
+			     netif_carrier_ok(adapter->netdev))) {
+			vmxnet3_tq_wake(tq, adapter);
+		}
+		spin_unlock(&tq->tx_lock);
+	}
+	return completed;
+}
+
+
+static void
+vmxnet3_tq_cleanup(struct vmxnet3_tx_queue *tq,
+		   struct vmxnet3_adapter *adapter)
+{
+	int i;
+
+	while (tq->tx_ring.next2comp != tq->tx_ring.next2fill) {
+		struct vmxnet3_tx_buf_info *tbi;
+		union Vmxnet3_GenericDesc *gdesc;
+
+		tbi = tq->buf_info + tq->tx_ring.next2comp;
+		gdesc = tq->tx_ring.base + tq->tx_ring.next2comp;
+
+		vmxnet3_unmap_tx_buf(tbi, adapter->pdev);
+		if (tbi->skb) {
+			dev_kfree_skb_any(tbi->skb);
+			tbi->skb = NULL;
+		}
+		vmxnet3_cmd_ring_adv_next2comp(&tq->tx_ring);
+	}
+
+	/* sanity check, verify all buffers are indeed unmapped and freed */
+	for (i = 0; i < tq->tx_ring.size; i++) {
+		BUG_ON(tq->buf_info[i].skb != NULL ||
+		       tq->buf_info[i].map_type != VMXNET3_MAP_NONE);
+	}
+
+	tq->tx_ring.gen = VMXNET3_INIT_GEN;
+	tq->tx_ring.next2fill = tq->tx_ring.next2comp = 0;
+
+	tq->comp_ring.gen = VMXNET3_INIT_GEN;
+	tq->comp_ring.next2proc = 0;
+}
+
+
+void
+vmxnet3_tq_destroy(struct vmxnet3_tx_queue *tq,
+		   struct vmxnet3_adapter *adapter)
+{
+	if (tq->tx_ring.base) {
+		pci_free_consistent(adapter->pdev, tq->tx_ring.size *
+				    sizeof(struct Vmxnet3_TxDesc),
+				    tq->tx_ring.base, tq->tx_ring.basePA);
+		tq->tx_ring.base = NULL;
+	}
+	if (tq->data_ring.base) {
+		pci_free_consistent(adapter->pdev, tq->data_ring.size *
+				    sizeof(struct Vmxnet3_TxDataDesc),
+				    tq->data_ring.base, tq->data_ring.basePA);
+		tq->data_ring.base = NULL;
+	}
+	if (tq->comp_ring.base) {
+		pci_free_consistent(adapter->pdev, tq->comp_ring.size *
+				    sizeof(struct Vmxnet3_TxCompDesc),
+				    tq->comp_ring.base, tq->comp_ring.basePA);
+		tq->comp_ring.base = NULL;
+	}
+	kfree(tq->buf_info);
+	tq->buf_info = NULL;
+}
+
+
+static void
+vmxnet3_tq_init(struct vmxnet3_tx_queue *tq,
+		struct vmxnet3_adapter *adapter)
+{
+	int i;
+
+	/* reset the tx ring contents to 0 and reset the tx ring states */
+	memset(tq->tx_ring.base, 0, tq->tx_ring.size *
+	       sizeof(struct Vmxnet3_TxDesc));
+	tq->tx_ring.next2fill = tq->tx_ring.next2comp = 0;
+	tq->tx_ring.gen = VMXNET3_INIT_GEN;
+
+	memset(tq->data_ring.base, 0, tq->data_ring.size *
+	       sizeof(struct Vmxnet3_TxDataDesc));
+
+	/* reset the tx comp ring contents to 0 and reset comp ring states */
+	memset(tq->comp_ring.base, 0, tq->comp_ring.size *
+	       sizeof(struct Vmxnet3_TxCompDesc));
+	tq->comp_ring.next2proc = 0;
+	tq->comp_ring.gen = VMXNET3_INIT_GEN;
+
+	/* reset the bookkeeping data */
+	memset(tq->buf_info, 0, sizeof(tq->buf_info[0]) * tq->tx_ring.size);
+	for (i = 0; i < tq->tx_ring.size; i++)
+		tq->buf_info[i].map_type = VMXNET3_MAP_NONE;
+
+	/* stats are not reset */
+}
+
+
+static int
+vmxnet3_tq_create(struct vmxnet3_tx_queue *tq,
+		  struct vmxnet3_adapter *adapter)
+{
+	BUG_ON(tq->tx_ring.base || tq->data_ring.base ||
+	       tq->comp_ring.base || tq->buf_info);
+
+	tq->tx_ring.base = pci_alloc_consistent(adapter->pdev, tq->tx_ring.size
+			   * sizeof(struct Vmxnet3_TxDesc),
+			   &tq->tx_ring.basePA);
+	if (!tq->tx_ring.base) {
+		printk(KERN_ERR "%s: failed to allocate tx ring\n",
+		       adapter->netdev->name);
+		goto err;
+	}
+
+	tq->data_ring.base = pci_alloc_consistent(adapter->pdev,
+			     tq->data_ring.size *
+			     sizeof(struct Vmxnet3_TxDataDesc),
+			     &tq->data_ring.basePA);
+	if (!tq->data_ring.base) {
+		printk(KERN_ERR "%s: failed to allocate data ring\n",
+		       adapter->netdev->name);
+		goto err;
+	}
+
+	tq->comp_ring.base = pci_alloc_consistent(adapter->pdev,
+			     tq->comp_ring.size *
+			     sizeof(struct Vmxnet3_TxCompDesc),
+			     &tq->comp_ring.basePA);
+	if (!tq->comp_ring.base) {
+		printk(KERN_ERR "%s: failed to allocate tx comp ring\n",
+		       adapter->netdev->name);
+		goto err;
+	}
+
+	tq->buf_info = kcalloc(tq->tx_ring.size, sizeof(tq->buf_info[0]),
+			       GFP_KERNEL);
+	if (!tq->buf_info) {
+		printk(KERN_ERR "%s: failed to allocate tx bufinfo\n",
+		       adapter->netdev->name);
+		goto err;
+	}
+
+	return 0;
+
+err:
+	vmxnet3_tq_destroy(tq, adapter);
+	return -ENOMEM;
+}
+
+
+/*
+ *    starting from ring->next2fill, allocate rx buffers for the given ring
+ *    of the rx queue and update the rx desc. stop after @num_to_alloc buffers
+ *    are allocated or allocation fails
+ */
+
+static int
+vmxnet3_rq_alloc_rx_buf(struct vmxnet3_rx_queue *rq, u32 ring_idx,
+			int num_to_alloc, struct vmxnet3_adapter *adapter)
+{
+	int num_allocated = 0;
+	struct vmxnet3_rx_buf_info *rbi_base = rq->buf_info[ring_idx];
+	struct vmxnet3_cmd_ring *ring = &rq->rx_ring[ring_idx];
+	u32 val;
+
+	while (num_allocated < num_to_alloc) {
+		struct vmxnet3_rx_buf_info *rbi;
+		union Vmxnet3_GenericDesc *gd;
+
+		rbi = rbi_base + ring->next2fill;
+		gd = ring->base + ring->next2fill;
+
+		if (rbi->buf_type == VMXNET3_RX_BUF_SKB) {
+			if (rbi->skb == NULL) {
+				rbi->skb = dev_alloc_skb(rbi->len +
+							 NET_IP_ALIGN);
+				if (unlikely(rbi->skb == NULL)) {
+					rq->stats.rx_buf_alloc_failure++;
+					break;
+				}
+				rbi->skb->dev = adapter->netdev;
+
+				skb_reserve(rbi->skb, NET_IP_ALIGN);
+				rbi->dma_addr = pci_map_single(adapter->pdev,
+						rbi->skb->data, rbi->len,
+						PCI_DMA_FROMDEVICE);
+			} else {
+				/* rx buffer skipped by the device */
+			}
+			val = VMXNET3_RXD_BTYPE_HEAD << VMXNET3_RXD_BTYPE_SHIFT;
+		} else {
+			BUG_ON(rbi->buf_type != VMXNET3_RX_BUF_PAGE ||
+			       rbi->len  != PAGE_SIZE);
+
+			if (rbi->page == NULL) {
+				rbi->page = alloc_page(GFP_ATOMIC);
+				if (unlikely(rbi->page == NULL)) {
+					rq->stats.rx_buf_alloc_failure++;
+					break;
+				}
+				rbi->dma_addr = pci_map_page(adapter->pdev,
+						rbi->page, 0, PAGE_SIZE,
+						PCI_DMA_FROMDEVICE);
+			} else {
+				/* rx buffers skipped by the device */
+			}
+			val = VMXNET3_RXD_BTYPE_BODY << VMXNET3_RXD_BTYPE_SHIFT;
+		}
+
+		BUG_ON(rbi->dma_addr == 0);
+		gd->rxd.addr = rbi->dma_addr;
+		gd->dword[2] = (ring->gen << VMXNET3_RXD_GEN_SHIFT) | val |
+				rbi->len;
+
+		num_allocated++;
+		vmxnet3_cmd_ring_adv_next2fill(ring);
+	}
+	rq->uncommitted[ring_idx] += num_allocated;
+
+	dprintk(KERN_ERR "alloc_rx_buf: %d allocated, next2fill %u, next2comp "
+		"%u, uncommited %u\n", num_allocated, ring->next2fill,
+		ring->next2comp, rq->uncommitted[ring_idx]);
+
+	/* so that the device can distinguish a full ring and an empty ring */
+	BUG_ON(num_allocated != 0 && ring->next2fill == ring->next2comp);
+
+	return num_allocated;
+}
+
+
+static void
+vmxnet3_append_frag(struct sk_buff *skb, struct Vmxnet3_RxCompDesc *rcd,
+		    struct vmxnet3_rx_buf_info *rbi)
+{
+	struct skb_frag_struct *frag = skb_shinfo(skb)->frags +
+		skb_shinfo(skb)->nr_frags;
+
+	BUG_ON(skb_shinfo(skb)->nr_frags >= MAX_SKB_FRAGS);
+
+	frag->page = rbi->page;
+	frag->page_offset = 0;
+	frag->size = rcd->len;
+	skb->data_len += frag->size;
+	skb_shinfo(skb)->nr_frags++;
+}
+
+
+static void
+vmxnet3_map_pkt(struct sk_buff *skb, struct vmxnet3_tx_ctx *ctx,
+		struct vmxnet3_tx_queue *tq, struct pci_dev *pdev,
+		struct vmxnet3_adapter *adapter)
+{
+	u32 dw2, len;
+	unsigned long buf_offset;
+	int i;
+	union Vmxnet3_GenericDesc *gdesc;
+	struct vmxnet3_tx_buf_info *tbi = NULL;
+
+	BUG_ON(ctx->copy_size > skb_headlen(skb));
+
+	/* use the previous gen bit for the SOP desc */
+	dw2 = (tq->tx_ring.gen ^ 0x1) << VMXNET3_TXD_GEN_SHIFT;
+
+	ctx->sop_txd = tq->tx_ring.base + tq->tx_ring.next2fill;
+	gdesc = ctx->sop_txd; /* both loops below can be skipped */
+
+	/* no need to map the buffer if headers are copied */
+	if (ctx->copy_size) {
+		ctx->sop_txd->txd.addr = tq->data_ring.basePA +
+					tq->tx_ring.next2fill *
+					sizeof(struct Vmxnet3_TxDataDesc);
+		ctx->sop_txd->dword[2] = dw2 | ctx->copy_size;
+		ctx->sop_txd->dword[3] = 0;
+
+		tbi = tq->buf_info + tq->tx_ring.next2fill;
+		tbi->map_type = VMXNET3_MAP_NONE;
+
+		dprintk(KERN_ERR "txd[%u]: 0x%Lx 0x%x 0x%x\n",
+			tq->tx_ring.next2fill, ctx->sop_txd->txd.addr,
+			ctx->sop_txd->dword[2], ctx->sop_txd->dword[3]);
+		vmxnet3_cmd_ring_adv_next2fill(&tq->tx_ring);
+
+		/* use the right gen for non-SOP desc */
+		dw2 = tq->tx_ring.gen << VMXNET3_TXD_GEN_SHIFT;
+	}
+
+	/* linear part can use multiple tx desc if it's big */
+	len = skb_headlen(skb) - ctx->copy_size;
+	buf_offset = ctx->copy_size;
+	while (len) {
+		u32 buf_size;
+
+		buf_size = len > VMXNET3_MAX_TX_BUF_SIZE ?
+			   VMXNET3_MAX_TX_BUF_SIZE : len;
+
+		tbi = tq->buf_info + tq->tx_ring.next2fill;
+		tbi->map_type = VMXNET3_MAP_SINGLE;
+		tbi->dma_addr = pci_map_single(adapter->pdev,
+				skb->data + buf_offset, buf_size,
+				PCI_DMA_TODEVICE);
+
+		tbi->len = buf_size; /* this automatically convert 2^14 to 0 */
+
+		gdesc = tq->tx_ring.base + tq->tx_ring.next2fill;
+		BUG_ON(gdesc->txd.gen == tq->tx_ring.gen);
+
+		gdesc->txd.addr = tbi->dma_addr;
+		gdesc->dword[2] = dw2 | buf_size;
+		gdesc->dword[3] = 0;
+
+		dprintk(KERN_ERR "txd[%u]: 0x%Lx 0x%x 0x%x\n",
+			tq->tx_ring.next2fill, gdesc->txd.addr,
+			gdesc->dword[2], gdesc->dword[3]);
+		vmxnet3_cmd_ring_adv_next2fill(&tq->tx_ring);
+		dw2 = tq->tx_ring.gen << VMXNET3_TXD_GEN_SHIFT;
+
+		len -= buf_size;
+		buf_offset += buf_size;
+	}
+
+	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
+		struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i];
+
+		tbi = tq->buf_info + tq->tx_ring.next2fill;
+		tbi->map_type = VMXNET3_MAP_PAGE;
+		tbi->dma_addr = pci_map_page(adapter->pdev, frag->page,
+					     frag->page_offset, frag->size,
+					     PCI_DMA_TODEVICE);
+
+		tbi->len = frag->size;
+
+		gdesc = tq->tx_ring.base + tq->tx_ring.next2fill;
+		BUG_ON(gdesc->txd.gen == tq->tx_ring.gen);
+
+		gdesc->txd.addr = tbi->dma_addr;
+		gdesc->dword[2] = dw2 | frag->size;
+		gdesc->dword[3] = 0;
+
+		dprintk(KERN_ERR "txd[%u]: 0x%llu %u %u\n",
+			tq->tx_ring.next2fill, gdesc->txd.addr,
+			gdesc->dword[2], gdesc->dword[3]);
+		vmxnet3_cmd_ring_adv_next2fill(&tq->tx_ring);
+		dw2 = tq->tx_ring.gen << VMXNET3_TXD_GEN_SHIFT;
+	}
+
+	ctx->eop_txd = gdesc;
+
+	/* set the last buf_info for the pkt */
+	tbi->skb = skb;
+	tbi->sop_idx = ctx->sop_txd - tq->tx_ring.base;
+}
+
+
+/*
+ *    parse and copy relevant protocol headers:
+ *      For a tso pkt, relevant headers are L2/3/4 including options
+ *      For a pkt requesting csum offloading, they are L2/3 and may include L4
+ *      if it's a TCP/UDP pkt
+ *
+ * Returns:
+ *    -1:  error happens during parsing
+ *     0:  protocol headers parsed, but too big to be copied
+ *     1:  protocol headers parsed and copied
+ *
+ * Other effects:
+ *    1. related *ctx fields are updated.
+ *    2. ctx->copy_size is # of bytes copied
+ *    3. the portion copied is guaranteed to be in the linear part
+ *
+ */
+static int
+vmxnet3_parse_and_copy_hdr(struct sk_buff *skb, struct vmxnet3_tx_queue *tq,
+			   struct vmxnet3_tx_ctx *ctx,
+			   struct vmxnet3_adapter *adapter)
+{
+	struct Vmxnet3_TxDataDesc *tdd;
+
+	if (ctx->mss) {
+		ctx->eth_ip_hdr_size = skb_transport_offset(skb);
+		ctx->l4_hdr_size = ((struct tcphdr *)
+				   skb_transport_header(skb))->doff * 4;
+		ctx->copy_size = ctx->eth_ip_hdr_size + ctx->l4_hdr_size;
+	} else {
+		unsigned int pull_size;
+
+		if (skb->ip_summed == CHECKSUM_PARTIAL) {
+			ctx->eth_ip_hdr_size = skb_transport_offset(skb);
+
+			if (ctx->ipv4) {
+				struct iphdr *iph = (struct iphdr *)
+						    skb_network_header(skb);
+				if (iph->protocol == IPPROTO_TCP) {
+					pull_size = ctx->eth_ip_hdr_size +
+						    sizeof(struct tcphdr);
+
+					if (unlikely(!pskb_may_pull(skb,
+								pull_size))) {
+						goto err;
+					}
+					ctx->l4_hdr_size = ((struct tcphdr *)
+					   skb_transport_header(skb))->doff * 4;
+				} else if (iph->protocol == IPPROTO_UDP) {
+					ctx->l4_hdr_size =
+							sizeof(struct udphdr);
+				} else {
+					ctx->l4_hdr_size = 0;
+				}
+			} else {
+				/* for simplicity, don't copy L4 headers */
+				ctx->l4_hdr_size = 0;
+			}
+			ctx->copy_size = ctx->eth_ip_hdr_size +
+					 ctx->l4_hdr_size;
+		} else {
+			ctx->eth_ip_hdr_size = 0;
+			ctx->l4_hdr_size = 0;
+			/* copy as much as allowed */
+			ctx->copy_size = min((unsigned int)VMXNET3_HDR_COPY_SIZE
+					     , skb_headlen(skb));
+		}
+
+		/* make sure headers are accessible directly */
+		if (unlikely(!pskb_may_pull(skb, ctx->copy_size)))
+			goto err;
+	}
+
+	if (unlikely(ctx->copy_size > VMXNET3_HDR_COPY_SIZE)) {
+		tq->stats.oversized_hdr++;
+		ctx->copy_size = 0;
+		return 0;
+	}
+
+	tdd = tq->data_ring.base + tq->tx_ring.next2fill;
+
+	memcpy(tdd->data, skb->data, ctx->copy_size);
+	dprintk(KERN_ERR "copy %u bytes to dataRing[%u]\n",
+		ctx->copy_size, tq->tx_ring.next2fill);
+	return 1;
+
+err:
+	return -1;
+}
+
+
+static void
+vmxnet3_prepare_tso(struct sk_buff *skb,
+		    struct vmxnet3_tx_ctx *ctx)
+{
+	struct tcphdr *tcph = (struct tcphdr *)skb_transport_header(skb);
+	if (ctx->ipv4) {
+		struct iphdr *iph = (struct iphdr *)skb_network_header(skb);
+		iph->check = 0;
+		tcph->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr, 0,
+						 IPPROTO_TCP, 0);
+	} else {
+		struct ipv6hdr *iph = (struct ipv6hdr *)skb_network_header(skb);
+		tcph->check = ~csum_ipv6_magic(&iph->saddr, &iph->daddr, 0,
+					       IPPROTO_TCP, 0);
+	}
+}
+
+
+/*
+ * Transmits a pkt thru a given tq
+ * Returns:
+ *    NETDEV_TX_OK:      descriptors are setup successfully
+ *    NETDEV_TX_OK:      error occured, the pkt is dropped
+ *    NETDEV_TX_BUSY:    tx ring is full, queue is stopped
+ *
+ * Side-effects:
+ *    1. tx ring may be changed
+ *    2. tq stats may be updated accordingly
+ *    3. shared->txNumDeferred may be updated
+ */
+
+static int
+vmxnet3_tq_xmit(struct sk_buff *skb, struct vmxnet3_tx_queue *tq,
+		struct vmxnet3_adapter *adapter, struct net_device *netdev)
+{
+	int ret;
+	u32 count;
+	unsigned long flags;
+	struct vmxnet3_tx_ctx ctx;
+	union Vmxnet3_GenericDesc *gdesc;
+
+	/* conservatively estimate # of descriptors to use */
+	count = VMXNET3_TXD_NEEDED(skb_headlen(skb)) +
+		skb_shinfo(skb)->nr_frags + 1;
+
+	ctx.ipv4 = (skb->protocol == __constant_ntohs(ETH_P_IP));
+
+	ctx.mss = skb_shinfo(skb)->gso_size;
+	if (ctx.mss) {
+		if (skb_header_cloned(skb)) {
+			if (unlikely(pskb_expand_head(skb, 0, 0,
+						      GFP_ATOMIC) != 0)) {
+				tq->stats.drop_tso++;
+				goto drop_pkt;
+			}
+			tq->stats.copy_skb_header++;
+		}
+		vmxnet3_prepare_tso(skb, &ctx);
+	} else {
+		if (unlikely(count > VMXNET3_MAX_TXD_PER_PKT)) {
+
+			/* non-tso pkts must not use more than
+			 * VMXNET3_MAX_TXD_PER_PKT entries
+			 */
+			if (skb_linearize(skb) != 0) {
+				tq->stats.drop_too_many_frags++;
+				goto drop_pkt;
+			}
+			tq->stats.linearized++;
+
+			/* recalculate the # of descriptors to use */
+			count = VMXNET3_TXD_NEEDED(skb_headlen(skb)) + 1;
+		}
+	}
+
+	ret = vmxnet3_parse_and_copy_hdr(skb, tq, &ctx, adapter);
+	if (ret >= 0) {
+		BUG_ON(ret <= 0 && ctx.copy_size != 0);
+		/* hdrs parsed, check against other limits */
+		if (ctx.mss) {
+			if (unlikely(ctx.eth_ip_hdr_size + ctx.l4_hdr_size >
+				     VMXNET3_MAX_TX_BUF_SIZE)) {
+				goto hdr_too_big;
+			}
+		} else {
+			if (skb->ip_summed == CHECKSUM_PARTIAL) {
+				if (unlikely(ctx.eth_ip_hdr_size +
+					     skb->csum_offset >
+					     VMXNET3_MAX_CSUM_OFFSET)) {
+					goto hdr_too_big;
+				}
+			}
+		}
+	} else {
+		tq->stats.drop_hdr_inspect_err++;
+		goto drop_pkt;
+	}
+
+	spin_lock_irqsave(&tq->tx_lock, flags);
+
+	if (count > vmxnet3_cmd_ring_desc_avail(&tq->tx_ring)) {
+		tq->stats.tx_ring_full++;
+		dprintk(KERN_ERR "tx queue stopped on %s, next2comp %u"
+			" next2fill %u\n", adapter->netdev->name,
+			tq->tx_ring.next2comp, tq->tx_ring.next2fill);
+
+		vmxnet3_tq_stop(tq, adapter);
+		spin_unlock_irqrestore(&tq->tx_lock, flags);
+		return NETDEV_TX_BUSY;
+	}
+
+	/* fill tx descs related to addr & len */
+	vmxnet3_map_pkt(skb, &ctx, tq, adapter->pdev, adapter);
+
+	/* setup the EOP desc */
+	ctx.eop_txd->dword[3] = VMXNET3_TXD_CQ | VMXNET3_TXD_EOP;
+
+	/* setup the SOP desc */
+	gdesc = ctx.sop_txd;
+	if (ctx.mss) {
+		gdesc->txd.hlen = ctx.eth_ip_hdr_size + ctx.l4_hdr_size;
+		gdesc->txd.om = VMXNET3_OM_TSO;
+		gdesc->txd.msscof = ctx.mss;
+		tq->shared->txNumDeferred += (skb->len - gdesc->txd.hlen +
+					     ctx.mss - 1) / ctx.mss;
+	} else {
+		if (skb->ip_summed == CHECKSUM_PARTIAL) {
+			gdesc->txd.hlen = ctx.eth_ip_hdr_size;
+			gdesc->txd.om = VMXNET3_OM_CSUM;
+			gdesc->txd.msscof = ctx.eth_ip_hdr_size +
+					    skb->csum_offset;
+		} else {
+			gdesc->txd.om = 0;
+			gdesc->txd.msscof = 0;
+		}
+		tq->shared->txNumDeferred++;
+	}
+
+	if (vlan_tx_tag_present(skb)) {
+		gdesc->txd.ti = 1;
+		gdesc->txd.tci = vlan_tx_tag_get(skb);
+	}
+
+	wmb();
+
+	/* finally flips the GEN bit of the SOP desc */
+	gdesc->dword[2] ^= VMXNET3_TXD_GEN;
+	dprintk(KERN_ERR "txd[%u]: SOP 0x%Lx 0x%x 0x%x\n",
+		(u32)((union Vmxnet3_GenericDesc *)ctx.sop_txd -
+		tq->tx_ring.base), gdesc->txd.addr, gdesc->dword[2],
+		gdesc->dword[3]);
+
+	spin_unlock_irqrestore(&tq->tx_lock, flags);
+
+	if (tq->shared->txNumDeferred >= tq->shared->txThreshold) {
+		tq->shared->txNumDeferred = 0;
+		VMXNET3_WRITE_BAR0_REG(adapter, VMXNET3_REG_TXPROD,
+				       tq->tx_ring.next2fill);
+	}
+	netdev->trans_start = jiffies;
+
+	return NETDEV_TX_OK;
+
+hdr_too_big:
+	tq->stats.drop_oversized_hdr++;
+drop_pkt:
+	tq->stats.drop_total++;
+	dev_kfree_skb(skb);
+	return NETDEV_TX_OK;
+}
+
+
+static netdev_tx_t
+vmxnet3_xmit_frame(struct sk_buff *skb, struct net_device *netdev)
+{
+	struct vmxnet3_adapter *adapter = netdev_priv(netdev);
+	struct vmxnet3_tx_queue *tq = &adapter->tx_queue;
+
+	return vmxnet3_tq_xmit(skb, tq, adapter, netdev);
+}
+
+
+static void
+vmxnet3_rx_csum(struct vmxnet3_adapter *adapter,
+		struct sk_buff *skb,
+		union Vmxnet3_GenericDesc *gdesc)
+{
+	if (!gdesc->rcd.cnc && adapter->rxcsum) {
+		/* typical case: TCP/UDP over IP and both csums are correct */
+		if ((gdesc->dword[3] & VMXNET3_RCD_CSUM_OK) ==
+							VMXNET3_RCD_CSUM_OK) {
+			skb->ip_summed = CHECKSUM_UNNECESSARY;
+			BUG_ON(!(gdesc->rcd.tcp || gdesc->rcd.udp));
+			BUG_ON(!(gdesc->rcd.v4  || gdesc->rcd.v6));
+			BUG_ON(gdesc->rcd.frg);
+		} else {
+			if (gdesc->rcd.csum) {
+				skb->csum = htons(gdesc->rcd.csum);
+				skb->ip_summed = CHECKSUM_PARTIAL;
+			} else {
+				skb->ip_summed = CHECKSUM_NONE;
+			}
+		}
+	} else {
+		skb->ip_summed = CHECKSUM_NONE;
+	}
+}
+
+
+static void
+vmxnet3_rx_error(struct vmxnet3_rx_queue *rq, struct Vmxnet3_RxCompDesc *rcd,
+		 struct vmxnet3_rx_ctx *ctx,  struct vmxnet3_adapter *adapter)
+{
+	rq->stats.drop_err++;
+	if (!rcd->fcs)
+		rq->stats.drop_fcs++;
+
+	rq->stats.drop_total++;
+
+	/*
+	 * We do not unmap and chain the rx buffer to the skb.
+	 * We basically pretend this buffer is not used and will be recycled
+	 * by vmxnet3_rq_alloc_rx_buf()
+	 */
+
+	/*
+	 * ctx->skb may be NULL if this is the first and the only one
+	 * desc for the pkt
+	 */
+	if (ctx->skb)
+		dev_kfree_skb_irq(ctx->skb);
+
+	ctx->skb = NULL;
+}
+
+
+static int
+vmxnet3_rq_rx_complete(struct vmxnet3_rx_queue *rq,
+		       struct vmxnet3_adapter *adapter, int quota)
+{
+	static u32 rxprod_reg[2] = {VMXNET3_REG_RXPROD, VMXNET3_REG_RXPROD2};
+	u32 num_rxd = 0;
+	struct Vmxnet3_RxCompDesc *rcd;
+	struct vmxnet3_rx_ctx *ctx = &rq->rx_ctx;
+
+	rcd = &rq->comp_ring.base[rq->comp_ring.next2proc].rcd;
+	while (rcd->gen == rq->comp_ring.gen) {
+		struct vmxnet3_rx_buf_info *rbi;
+		struct sk_buff *skb;
+		int num_to_alloc;
+		struct Vmxnet3_RxDesc *rxd;
+		u32 idx, ring_idx;
+
+		if (num_rxd >= quota) {
+			/* we may stop even before we see the EOP desc of
+			 * the current pkt
+			 */
+			break;
+		}
+		num_rxd++;
+
+		idx = rcd->rxdIdx;
+		ring_idx = rcd->rqID == rq->qid ? 0 : 1;
+
+		rxd = &rq->rx_ring[ring_idx].base[idx].rxd;
+		rbi = rq->buf_info[ring_idx] + idx;
+
+		BUG_ON(rxd->addr != rbi->dma_addr || rxd->len != rbi->len);
+
+		if (unlikely(rcd->eop && rcd->err)) {
+			vmxnet3_rx_error(rq, rcd, ctx, adapter);
+			goto rcd_done;
+		}
+
+		if (rcd->sop) { /* first buf of the pkt */
+			BUG_ON(rxd->btype != VMXNET3_RXD_BTYPE_HEAD ||
+			       rcd->rqID != rq->qid);
+
+			BUG_ON(rbi->buf_type != VMXNET3_RX_BUF_SKB);
+			BUG_ON(ctx->skb != NULL || rbi->skb == NULL);
+
+			if (unlikely(rcd->len == 0)) {
+				/* Pretend the rx buffer is skipped. */
+				BUG_ON(!(rcd->sop && rcd->eop));
+				dprintk(KERN_ERR "rxRing[%u][%u] 0 length\n",
+					ring_idx, idx);
+				goto rcd_done;
+			}
+
+			ctx->skb = rbi->skb;
+			rbi->skb = NULL;
+
+			pci_unmap_single(adapter->pdev, rbi->dma_addr, rbi->len,
+					 PCI_DMA_FROMDEVICE);
+
+			skb_put(ctx->skb, rcd->len);
+		} else {
+			BUG_ON(ctx->skb == NULL);
+			/* non SOP buffer must be type 1 in most cases */
+			if (rbi->buf_type == VMXNET3_RX_BUF_PAGE) {
+				BUG_ON(rxd->btype != VMXNET3_RXD_BTYPE_BODY);
+
+				if (rcd->len) {
+					pci_unmap_page(adapter->pdev,
+						       rbi->dma_addr, rbi->len,
+						       PCI_DMA_FROMDEVICE);
+
+					vmxnet3_append_frag(ctx->skb, rcd, rbi);
+					rbi->page = NULL;
+				}
+			} else {
+				/*
+				 * The only time a non-SOP buffer is type 0 is
+				 * when it's EOP and error flag is raised, which
+				 * has already been handled.
+				 */
+				BUG_ON(true);
+			}
+		}
+
+		skb = ctx->skb;
+		if (rcd->eop) {
+			skb->len += skb->data_len;
+			skb->truesize += skb->data_len;
+
+			vmxnet3_rx_csum(adapter, skb,
+					(union Vmxnet3_GenericDesc *)rcd);
+			skb->protocol = eth_type_trans(skb, adapter->netdev);
+
+			if (unlikely(adapter->vlan_grp && rcd->ts)) {
+				vlan_hwaccel_receive_skb(skb,
+						adapter->vlan_grp, rcd->tci);
+			} else {
+				netif_receive_skb(skb);
+			}
+
+			adapter->netdev->last_rx = jiffies;
+			ctx->skb = NULL;
+		}
+
+rcd_done:
+		/* device may skip some rx descs */
+		rq->rx_ring[ring_idx].next2comp = idx;
+		VMXNET3_INC_RING_IDX_ONLY(rq->rx_ring[ring_idx].next2comp,
+					  rq->rx_ring[ring_idx].size);
+
+		/* refill rx buffers frequently to avoid starving the h/w */
+		num_to_alloc = vmxnet3_cmd_ring_desc_avail(rq->rx_ring +
+							   ring_idx);
+		if (unlikely(num_to_alloc > VMXNET3_RX_ALLOC_THRESHOLD(rq,
+							ring_idx, adapter))) {
+			vmxnet3_rq_alloc_rx_buf(rq, ring_idx, num_to_alloc,
+						adapter);
+
+			/* if needed, update the register */
+			if (unlikely(rq->shared->updateRxProd)) {
+				VMXNET3_WRITE_BAR0_REG(adapter,
+					rxprod_reg[ring_idx] + rq->qid * 8,
+					rq->rx_ring[ring_idx].next2fill);
+				rq->uncommitted[ring_idx] = 0;
+			}
+		}
+
+		vmxnet3_comp_ring_adv_next2proc(&rq->comp_ring);
+		rcd = &rq->comp_ring.base[rq->comp_ring.next2proc].rcd;
+	}
+
+	return num_rxd;
+}
+
+
+static void
+vmxnet3_rq_cleanup(struct vmxnet3_rx_queue *rq,
+		   struct vmxnet3_adapter *adapter)
+{
+	u32 i, ring_idx;
+	struct Vmxnet3_RxDesc *rxd;
+
+	for (ring_idx = 0; ring_idx < 2; ring_idx++) {
+		for (i = 0; i < rq->rx_ring[ring_idx].size; i++) {
+			rxd = &rq->rx_ring[ring_idx].base[i].rxd;
+
+			if (rxd->btype == VMXNET3_RXD_BTYPE_HEAD &&
+					rq->buf_info[ring_idx][i].skb) {
+				pci_unmap_single(adapter->pdev, rxd->addr,
+						 rxd->len, PCI_DMA_FROMDEVICE);
+				dev_kfree_skb(rq->buf_info[ring_idx][i].skb);
+				rq->buf_info[ring_idx][i].skb = NULL;
+			} else if (rxd->btype == VMXNET3_RXD_BTYPE_BODY &&
+					rq->buf_info[ring_idx][i].page) {
+				pci_unmap_page(adapter->pdev, rxd->addr,
+					       rxd->len, PCI_DMA_FROMDEVICE);
+				put_page(rq->buf_info[ring_idx][i].page);
+				rq->buf_info[ring_idx][i].page = NULL;
+			}
+		}
+
+		rq->rx_ring[ring_idx].gen = VMXNET3_INIT_GEN;
+		rq->rx_ring[ring_idx].next2fill =
+					rq->rx_ring[ring_idx].next2comp = 0;
+		rq->uncommitted[ring_idx] = 0;
+	}
+
+	rq->comp_ring.gen = VMXNET3_INIT_GEN;
+	rq->comp_ring.next2proc = 0;
+}
+
+
+void vmxnet3_rq_destroy(struct vmxnet3_rx_queue *rq,
+			struct vmxnet3_adapter *adapter)
+{
+	int i;
+	int j;
+
+	/* all rx buffers must have already been freed */
+	for (i = 0; i < 2; i++) {
+		if (rq->buf_info[i]) {
+			for (j = 0; j < rq->rx_ring[i].size; j++)
+				BUG_ON(rq->buf_info[i][j].page != NULL);
+		}