[PATCH 4/7] cxgb4: Add packet queues and packet DMA code

[PATCH 4/7] cxgb4: Add packet queues and packet DMA code

[Dimitris Michailidis]

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
---
 drivers/net/cxgb4/sge.c | 2461 +++++++++++++++++++++++++++++++++++++++++++++++
 1 files changed, 2461 insertions(+), 0 deletions(-)
 create mode 100644 drivers/net/cxgb4/sge.c

diff --git a/drivers/net/cxgb4/sge.c b/drivers/net/cxgb4/sge.c
new file mode 100644
index 0000000..34617f8
--- /dev/null
+++ b/drivers/net/cxgb4/sge.c
@@ -0,0 +1,2461 @@
+/*
+ * This file is part of the Chelsio T4 Ethernet driver for Linux.
+ *
+ * Copyright (c) 2003-2010 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/skbuff.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/if_vlan.h>
+#include <linux/ip.h>
+#include <linux/dma-mapping.h>
+#include <net/ipv6.h>
+#include <net/tcp.h>
+#include "cxgb4.h"
+#include "t4_regs.h"
+#include "t4_msg.h"
+#include "t4fw_api.h"
+
+/*
+ * Rx buffer size.  We use largish buffers if possible but settle for single
+ * pages under memory shortage.
+ */
+#if PAGE_SHIFT >= 16
+# define FL_PG_ORDER 0
+#else
+# define FL_PG_ORDER (16 - PAGE_SHIFT)
+#endif
+
+/* RX_PULL_LEN should be <= RX_COPY_THRES */
+#define RX_COPY_THRES    256
+#define RX_PULL_LEN      128
+
+/*
+ * Main body length for sk_buffs used for Rx Ethernet packets with fragments.
+ * Should be >= RX_PULL_LEN but possibly bigger to give pskb_may_pull some room.
+ */
+#define RX_PKT_SKB_LEN   512
+
+/* Ethernet header padding prepended to RX_PKTs */
+#define RX_PKT_PAD 2
+
+/*
+ * Max number of Tx descriptors we clean up at a time.  Should be modest as
+ * freeing skbs isn't cheap and it happens while holding locks.  We just need
+ * to free packets faster than they arrive, we eventually catch up and keep
+ * the amortized cost reasonable.  Must be >= 2 * TXQ_STOP_THRES.
+ */
+#define MAX_TX_RECLAIM 16
+
+/*
+ * Max number of Rx buffers we replenish at a time.  Again keep this modest,
+ * allocating buffers isn't cheap either.
+ */
+#define MAX_RX_REFILL 16U
+
+/*
+ * Period of the Rx queue check timer.  This timer is infrequent as it has
+ * something to do only when the system experiences severe memory shortage.
+ */
+#define RX_QCHECK_PERIOD (HZ / 2)
+
+/*
+ * Period of the Tx queue check timer.
+ */
+#define TX_QCHECK_PERIOD (HZ / 2)
+
+/*
+ * Timer index used when backing off due to memory shortage.
+ */
+#define NOMEM_TMR_IDX (SGE_NTIMERS - 1)
+
+/*
+ * An FL with <= FL_STARVE_THRES buffers is starving and a periodic timer will
+ * attempt to refill it.
+ */
+#define FL_STARVE_THRES 4
+
+/*
+ * Suspend an Ethernet Tx queue with fewer available descriptors than this.
+ * This is the same as calc_tx_descs() for a TSO packet with
+ * nr_frags == MAX_SKB_FRAGS.
+ */
+#define ETHTXQ_STOP_THRES \
+	(1 + DIV_ROUND_UP((3 * MAX_SKB_FRAGS) / 2 + (MAX_SKB_FRAGS & 1), 8))
+
+/*
+ * Suspension threshold for non-Ethernet Tx queues.  We require enough room
+ * for a full sized WR.
+ */
+#define TXQ_STOP_THRES (SGE_MAX_WR_LEN / sizeof(struct tx_desc))
+
+/*
+ * Max Tx descriptor space we allow for an Ethernet packet to be inlined
+ * into a WR.
+ */
+#define MAX_IMM_TX_PKT_LEN 128
+
+/*
+ * Max size of a WR sent through a control Tx queue.
+ */
+#define MAX_CTRL_WR_LEN SGE_MAX_WR_LEN
+
+enum {
+	/* packet alignment in FL buffers */
+	FL_ALIGN = L1_CACHE_BYTES < 32 ? 32 : L1_CACHE_BYTES,
+	/* egress status entry size */
+	STAT_LEN = L1_CACHE_BYTES > 64 ? 128 : 64
+};
+
+struct tx_sw_desc {                /* SW state per Tx descriptor */
+	struct sk_buff *skb;
+	struct ulptx_sgl *sgl;
+};
+
+struct rx_sw_desc {                /* SW state per Rx descriptor */
+	struct page *page;
+	dma_addr_t dma_addr;
+};
+
+/*
+ * The low bits of rx_sw_desc.dma_addr have special meaning.
+ */
+enum {
+	RX_LARGE_BUF    = 1 << 0, /* buffer is larger than PAGE_SIZE */
+	RX_UNMAPPED_BUF = 1 << 1, /* buffer is not mapped */
+};
+
+static inline dma_addr_t get_buf_addr(const struct rx_sw_desc *d)
+{
+	return d->dma_addr & ~(dma_addr_t)(RX_LARGE_BUF | RX_UNMAPPED_BUF);
+}
+
+static inline bool is_buf_mapped(const struct rx_sw_desc *d)
+{
+	return !(d->dma_addr & RX_UNMAPPED_BUF);
+}
+
+/**
+ *	txq_avail - return the number of available slots in a Tx queue
+ *	@q: the Tx queue
+ *
+ *	Returns the number of descriptors in a Tx queue available to write new
+ *	packets.
+ */
+static inline unsigned int txq_avail(const struct sge_txq *q)
+{
+	return q->size - 1 - q->in_use;
+}
+
+/**
+ *	fl_cap - return the capacity of a free-buffer list
+ *	@fl: the FL
+ *
+ *	Returns the capacity of a free-buffer list.  The capacity is less than
+ *	the size because one descriptor needs to be left unpopulated, otherwise
+ *	HW will think the FL is empty.
+ */
+static inline unsigned int fl_cap(const struct sge_fl *fl)
+{
+	return fl->size - 8;   /* 1 descriptor = 8 buffers */
+}
+
+static inline bool fl_starving(const struct sge_fl *fl)
+{
+	return fl->avail - fl->pend_cred <= FL_STARVE_THRES;
+}
+
+static int map_skb(struct device *dev, const struct sk_buff *skb,
+		   dma_addr_t *addr)
+{
+	const skb_frag_t *fp, *end;
+	const struct skb_shared_info *si;
+
+	*addr = dma_map_single(dev, skb->data, skb_headlen(skb), DMA_TO_DEVICE);
+	if (dma_mapping_error(dev, *addr))
+		goto out_err;
+
+	si = skb_shinfo(skb);
+	end = &si->frags[si->nr_frags];
+
+	for (fp = si->frags; fp < end; fp++) {
+		*++addr = dma_map_page(dev, fp->page, fp->page_offset, fp->size,
+				       DMA_TO_DEVICE);
+		if (dma_mapping_error(dev, *addr))
+			goto unwind;
+	}
+	return 0;
+
+unwind:
+	while (fp-- > si->frags)
+		dma_unmap_page(dev, *--addr, fp->size, DMA_TO_DEVICE);
+
+	dma_unmap_single(dev, addr[-1], skb_headlen(skb), DMA_TO_DEVICE);
+out_err:
+	return -ENOMEM;
+}
+
+static void unmap_skb(struct device *dev, const struct sk_buff *skb,
+		      const dma_addr_t *addr)
+{
+	const skb_frag_t *fp, *end;
+	const struct skb_shared_info *si;
+
+	dma_unmap_single(dev, *addr++, skb_headlen(skb), DMA_TO_DEVICE);
+
+	si = skb_shinfo(skb);
+	end = &si->frags[si->nr_frags];
+	for (fp = si->frags; fp < end; fp++)
+		dma_unmap_page(dev, *addr++, fp->size, DMA_TO_DEVICE);
+}
+
+static void unmap_sgl(struct device *dev, const struct sk_buff *skb,
+		      const struct ulptx_sgl *sgl, const struct sge_txq *q)
+{
+	const struct ulptx_sge_pair *p;
+	unsigned int nfrags = skb_shinfo(skb)->nr_frags;
+
+	if (likely(skb_headlen(skb)))
+		dma_unmap_single(dev, be64_to_cpu(sgl->addr0), ntohl(sgl->len0),
+				 DMA_TO_DEVICE);
+	else {
+		dma_unmap_page(dev, be64_to_cpu(sgl->addr0), ntohl(sgl->len0),
+			       DMA_TO_DEVICE);
+		nfrags--;
+	}
+
+	/*
+	 * the complexity below is because of the possibility of a wrap-around
+	 * in the middle of an SGL
+	 */
+	for (p = sgl->sge; nfrags >= 2; nfrags -= 2) {
+		if (likely((u8 *)(p + 1) <= (u8 *)q->stat)) {
+unmap:			dma_unmap_page(dev, be64_to_cpu(p->addr[0]),
+				       ntohl(p->len[0]), DMA_TO_DEVICE);
+			dma_unmap_page(dev, be64_to_cpu(p->addr[1]),
+				       ntohl(p->len[1]), DMA_TO_DEVICE);
+			p++;
+		} else if ((u8 *)p == (u8 *)q->stat) {
+			p = (const struct ulptx_sge_pair *)q->desc;
+			goto unmap;
+		} else if ((u8 *)p + 8 == (u8 *)q->stat) {
+			const __be64 *addr = (const __be64 *)q->desc;
+
+			dma_unmap_page(dev, be64_to_cpu(addr[0]),
+				       ntohl(p->len[0]), DMA_TO_DEVICE);
+			dma_unmap_page(dev, be64_to_cpu(addr[1]),
+				       ntohl(p->len[1]), DMA_TO_DEVICE);
+			p = (const struct ulptx_sge_pair *)&addr[2];
+		} else {
+			const __be64 *addr = (const __be64 *)q->desc;
+
+			dma_unmap_page(dev, be64_to_cpu(p->addr[0]),
+				       ntohl(p->len[0]), DMA_TO_DEVICE);
+			dma_unmap_page(dev, be64_to_cpu(addr[0]),
+				       ntohl(p->len[1]), DMA_TO_DEVICE);
+			p = (const struct ulptx_sge_pair *)&addr[1];
+		}
+	}
+	if (nfrags) {
+		__be64 addr;
+
+		if ((u8 *)p == (u8 *)q->stat)
+			p = (const struct ulptx_sge_pair *)q->desc;
+		addr = (u8 *)p + 16 <= (u8 *)q->stat ? p->addr[0] :
+						       *(const __be64 *)q->desc;
+		dma_unmap_page(dev, be64_to_cpu(addr), ntohl(p->len[0]),
+			       DMA_TO_DEVICE);
+	}
+}
+
+/**
+ *	need_skb_unmap - does the platform need unmapping of sk_buffs?
+ *
+ *	Returns true if the platfrom needs sk_buff unmapping.  The compiler
+ *	optimizes away unecessary code if this returns true.
+ */
+static inline int need_skb_unmap(void)
+{
+	/*
+	 * This structure is used to tell if the platfrom needs buffer
+	 * unmapping by checking if DECLARE_PCI_UNMAP_ADDR defines anything.
+	 */
+	struct dummy {
+		DECLARE_PCI_UNMAP_ADDR(addr);
+	};
+
+	return sizeof(struct dummy) != 0;
+}
+
+/**
+ *	free_tx_desc - reclaims Tx descriptors and their buffers
+ *	@adapter: the adapter
+ *	@q: the Tx queue to reclaim descriptors from
+ *	@n: the number of descriptors to reclaim
+ *	@unmap: whether the buffers should be unmapped for DMA
+ *
+ *	Reclaims Tx descriptors from an SGE Tx queue and frees the associated
+ *	Tx buffers.  Called with the Tx queue lock held.
+ */
+static void free_tx_desc(struct adapter *adap, struct sge_txq *q,
+			 unsigned int n, bool unmap)
+{
+	struct tx_sw_desc *d;
+	unsigned int cidx = q->cidx;
+	struct device *dev = adap->pdev_dev;
+
+	const int need_unmap = need_skb_unmap() && unmap;
+
+	d = &q->sdesc[cidx];
+	while (n--) {
+		if (d->skb) {                       /* an SGL is present */
+			if (need_unmap)
+				unmap_sgl(dev, d->skb, d->sgl, q);
+			kfree_skb(d->skb);
+			d->skb = NULL;
+		}
+		++d;
+		if (++cidx == q->size) {
+			cidx = 0;
+			d = q->sdesc;
+		}
+	}
+	q->cidx = cidx;
+}
+
+/*
+ * Return the number of reclaimable descriptors in a Tx queue.
+ */
+static inline int reclaimable(const struct sge_txq *q)
+{
+	int hw_cidx = ntohs(q->stat->cidx);
+	hw_cidx -= q->cidx;
+	return hw_cidx < 0 ? hw_cidx + q->size : hw_cidx;
+}
+
+/**
+ *	reclaim_completed_tx - reclaims completed Tx descriptors
+ *	@adap: the adapter
+ *	@q: the Tx queue to reclaim completed descriptors from
+ *	@unmap: whether the buffers should be unmapped for DMA
+ *
+ *	Reclaims Tx descriptors that the SGE has indicated it has processed,
+ *	and frees the associated buffers if possible.  Called with the Tx
+ *	queue locked.
+ */
+static inline void reclaim_completed_tx(struct adapter *adap, struct sge_txq *q,
+					bool unmap)
+{
+	int avail = reclaimable(q);
+
+	if (avail) {
+		/*
+		 * Limit the amount of clean up work we do at a time to keep
+		 * the Tx lock hold time O(1).
+		 */
+		if (avail > MAX_TX_RECLAIM)
+			avail = MAX_TX_RECLAIM;
+
+		free_tx_desc(adap, q, avail, unmap);
+		q->in_use -= avail;
+	}
+}
+
+static inline int get_buf_size(const struct rx_sw_desc *d)
+{
+#if FL_PG_ORDER > 0
+	return (d->dma_addr & RX_LARGE_BUF) ? (PAGE_SIZE << FL_PG_ORDER) :
+					      PAGE_SIZE;
+#else
+	return PAGE_SIZE;
+#endif
+}
+
+/**
+ *	free_rx_bufs - free the Rx buffers on an SGE free list
+ *	@adap: the adapter
+ *	@q: the SGE free list to free buffers from
+ *	@n: how many buffers to free
+ *
+ *	Release the next @n buffers on an SGE free-buffer Rx queue.   The
+ *	buffers must be made inaccessible to HW before calling this function.
+ */
+static void free_rx_bufs(struct adapter *adap, struct sge_fl *q, int n)
+{
+	while (n--) {
+		struct rx_sw_desc *d = &q->sdesc[q->cidx];
+
+		if (is_buf_mapped(d))
+			dma_unmap_page(adap->pdev_dev, get_buf_addr(d),
+				       get_buf_size(d), PCI_DMA_FROMDEVICE);
+		put_page(d->page);
+		d->page = NULL;
+		if (++q->cidx == q->size)
+			q->cidx = 0;
+		q->avail--;
+	}
+}
+
+/**
+ *	unmap_rx_buf - unmap the current Rx buffer on an SGE free list
+ *	@adap: the adapter
+ *	@q: the SGE free list
+ *
+ *	Unmap the current buffer on an SGE free-buffer Rx queue.   The
+ *	buffer must be made inaccessible to HW before calling this function.
+ *
+ *	This is similar to @free_rx_bufs above but does not free the buffer.
+ *	Do note that the FL still loses any further access to the buffer.
+ */
+static void unmap_rx_buf(struct adapter *adap, struct sge_fl *q)
+{
+	struct rx_sw_desc *d = &q->sdesc[q->cidx];
+
+	if (is_buf_mapped(d))
+		dma_unmap_page(adap->pdev_dev, get_buf_addr(d),
+			       get_buf_size(d), PCI_DMA_FROMDEVICE);
+	d->page = NULL;
+	if (++q->cidx == q->size)
+		q->cidx = 0;
+	q->avail--;
+}
+
+static inline void ring_fl_db(struct adapter *adap, struct sge_fl *q)
+{
+	if (q->pend_cred >= 8) {
+		t4_write_reg(adap, MYPF_REG(SGE_PF_KDOORBELL), DBPRIO |
+			     QID(q->cntxt_id) | PIDX(q->pend_cred / 8));
+		q->pend_cred &= 7;
+	}
+}
+
+static inline void set_rx_sw_desc(struct rx_sw_desc *sd, struct page *pg,
+				  dma_addr_t mapping)
+{
+	sd->page = pg;
+	sd->dma_addr = mapping;      /* includes size low bits */
+}
+
+/**
+ *	refill_fl - refill an SGE Rx buffer ring
+ *	@adap: the adapter
+ *	@q: the ring to refill
+ *	@n: the number of new buffers to allocate
+ *	@gfp: the gfp flags for the allocations
+ *
+ *	(Re)populate an SGE free-buffer queue with up to @n new packet buffers,
+ *	allocated with the supplied gfp flags.  The caller must assure that
+ *	@n does not exceed the queue's capacity.  Returns the number of buffers
+ *	allocated.
+ */
+static unsigned int refill_fl(struct adapter *adap, struct sge_fl *q, int n,
+			      gfp_t gfp)
+{
+	struct page *pg;
+	dma_addr_t mapping;
+	unsigned int cred = q->avail;
+	__be64 *d = &q->desc[q->pidx];
+	struct rx_sw_desc *sd = &q->sdesc[q->pidx];
+
+	gfp |= __GFP_NOWARN;         /* failures are expected */
+
+#if FL_PG_ORDER > 0
+	/*
+	 * Prefer large buffers
+	 */
+	while (n) {
+		pg = alloc_pages(gfp | __GFP_COMP, FL_PG_ORDER);
+		if (unlikely(!pg)) {
+			q->large_alloc_failed++;
+			break;       /* fall back to single pages */
+		}
+
+		mapping = dma_map_page(adap->pdev_dev, pg, 0,
+				       PAGE_SIZE << FL_PG_ORDER,
+				       PCI_DMA_FROMDEVICE);
+		if (unlikely(dma_mapping_error(adap->pdev_dev, mapping))) {
+			__free_pages(pg, FL_PG_ORDER);
+			goto out;   /* do not try small pages for this error */
+		}
+		mapping |= RX_LARGE_BUF;
+		*d++ = cpu_to_be64(mapping);
+
+		set_rx_sw_desc(sd, pg, mapping);
+		sd++;
+
+		q->avail++;
+		if (++q->pidx == q->size) {
+			q->pidx = 0;
+			sd = q->sdesc;
+			d = q->desc;
+		}
+		n--;
+	}
+#endif
+
+	while (n--) {
+		pg = __netdev_alloc_page(adap->port[0], gfp);
+		if (unlikely(!pg)) {
+			q->alloc_failed++;
+			break;
+		}
+
+		mapping = dma_map_page(adap->pdev_dev, pg, 0, PAGE_SIZE,
+				       PCI_DMA_FROMDEVICE);
+		if (unlikely(dma_mapping_error(adap->pdev_dev, mapping))) {
+			netdev_free_page(adap->port[0], pg);
+			break;
+		}
+		*d++ = cpu_to_be64(mapping);
+
+		set_rx_sw_desc(sd, pg, mapping);
+		sd++;
+
+		q->avail++;
+		if (++q->pidx == q->size) {
+			q->pidx = 0;
+			sd = q->sdesc;
+			d = q->desc;
+		}
+	}
+
+out:	cred = q->avail - cred;
+	q->pend_cred += cred;
+	ring_fl_db(adap, q);
+	return cred;
+}
+
+static inline void __refill_fl(struct adapter *adap, struct sge_fl *fl)
+{
+	refill_fl(adap, fl, min(MAX_RX_REFILL, fl_cap(fl) - fl->avail),
+		  GFP_ATOMIC);
+}
+
+/**
+ *	replenish_fl - refill an SGE Rx buffer ring and check for starvation
+ *	@adap: the adapter
+ *	@fl: the ring to refill
+ *	@gfp: the gfp flags for the allocations
+ *
+ *	Attempt to refill an FL to capacity.  If afterwards the queue is found
+ *	critically low mark it as starving in the bitmap of starving FLs.
+ */
+static void replenish_fl(struct adapter *adap, struct sge_fl *fl, gfp_t gfp)
+{
+	refill_fl(adap, fl, fl_cap(fl) - fl->avail, gfp);
+	if (unlikely(fl_starving(fl)))
+		set_bit(fl->cntxt_id, adap->sge.starving_fl);
+}
+
+/**
+ *	alloc_ring - allocate resources for an SGE descriptor ring
+ *	@dev: the PCI device's core device
+ *	@nelem: the number of descriptors
+ *	@elem_size: the size of each descriptor
+ *	@sw_size: the size of the SW state associated with each ring element
+ *	@phys: the physical address of the allocated ring
+ *	@metadata: address of the array holding the SW state for the ring
+ *	@stat_size: extra space in HW ring for status information
+ *
+ *	Allocates resources for an SGE descriptor ring, such as Tx queues,
+ *	free buffer lists, or response queues.  Each SGE ring requires
+ *	space for its HW descriptors plus, optionally, space for the SW state
+ *	associated with each HW entry (the metadata).  The function returns
+ *	three values: the virtual address for the HW ring (the return value
+ *	of the function), the bus address of the HW ring, and the address
+ *	of the SW ring.
+ */
+static void *alloc_ring(struct device *dev, size_t nelem, size_t elem_size,
+			size_t sw_size, dma_addr_t *phys, void *metadata,
+			size_t stat_size)
+{
+	size_t len = nelem * elem_size + stat_size;
+	void *s = NULL;
+	void *p = dma_alloc_coherent(dev, len, phys, GFP_KERNEL);
+
+	if (!p)
+		return NULL;
+	if (sw_size) {
+		s = kcalloc(nelem, sw_size, GFP_KERNEL);
+
+		if (!s) {
+			dma_free_coherent(dev, len, p, *phys);
+			return NULL;
+		}
+	}
+	if (metadata)
+		*(void **)metadata = s;
+	memset(p, 0, len);
+	return p;
+}
+
+/**
+ *	sgl_len - calculates the size of an SGL of the given capacity
+ *	@n: the number of SGL entries
+ *
+ *	Calculates the number of flits needed for a scatter/gather list that
+ *	can hold the given number of entries.
+ */
+static inline unsigned int sgl_len(unsigned int n)
+{
+	n--;
+	return (3 * n) / 2 + (n & 1) + 2;
+}
+
+/**
+ *	flits_to_desc - returns the num of Tx descriptors for the given flits
+ *	@n: the number of flits
+ *
+ *	Returns the number of Tx descriptors needed for the supplied number
+ *	of flits.
+ */
+static inline unsigned int flits_to_desc(unsigned int n)
+{
+	BUG_ON(n > SGE_MAX_WR_LEN / 8);
+	return DIV_ROUND_UP(n, 8);
+}
+
+/**
+ *	is_eth_imm - can an Ethernet packet be sent as immediate data?
+ *	@skb: the packet
+ *
+ *	Returns whether an Ethernet packet is small enough to fit as
+ *	immediate data.
+ */
+static inline int is_eth_imm(const struct sk_buff *skb)
+{
+	return skb->len <= MAX_IMM_TX_PKT_LEN - sizeof(struct cpl_tx_pkt);
+}
+
+/**
+ *	calc_tx_flits - calculate the number of flits for a packet Tx WR
+ *	@skb: the packet
+ *
+ * 	Returns the number of flits needed for a Tx WR for the given Ethernet
+ * 	packet, including the needed WR and CPL headers.
+ */
+static inline unsigned int calc_tx_flits(const struct sk_buff *skb)
+{
+	unsigned int flits;
+
+	if (is_eth_imm(skb))
+		return DIV_ROUND_UP(skb->len + sizeof(struct cpl_tx_pkt), 8);
+
+	flits = sgl_len(skb_shinfo(skb)->nr_frags + 1) + 4;
+	if (skb_shinfo(skb)->gso_size)
+		flits += 2;
+	return flits;
+}
+
+/**
+ *	calc_tx_descs - calculate the number of Tx descriptors for a packet
+ *	@skb: the packet
+ *
+ * 	Returns the number of Tx descriptors needed for the given Ethernet
+ * 	packet, including the needed WR and CPL headers.
+ */
+static inline unsigned int calc_tx_descs(const struct sk_buff *skb)
+{
+	return flits_to_desc(calc_tx_flits(skb));
+}
+
+/**
+ *	write_sgl - populate a scatter/gather list for a packet
+ *	@skb: the packet
+ *	@q: the Tx queue we are writing into
+ *	@sgl: starting location for writing the SGL
+ *	@end: points right after the end of the SGL
+ *	@start: start offset into skb main-body data to include in the SGL
+ *	@addr: the list of bus addresses for the SGL elements
+ *
+ *	Generates a gather list for the buffers that make up a packet.
+ *	The caller must provide adequate space for the SGL that will be written.
+ *	The SGL includes all of the packet's page fragments and the data in its
+ *	main body except for the first @start bytes.  @sgl must be 16-byte
+ *	aligned and within a Tx descriptor with available space.  @end points
+ *	right after the end of the SGL but does not account for any potential
+ *	wrap around, i.e., @end > @sgl.
+ */
+static void write_sgl(const struct sk_buff *skb, struct sge_txq *q,
+		      struct ulptx_sgl *sgl, u64 *end, unsigned int start,
+		      const dma_addr_t *addr)
+{
+	unsigned int i, len;
+	struct ulptx_sge_pair *to;
+	const struct skb_shared_info *si = skb_shinfo(skb);
+	unsigned int nfrags = si->nr_frags;
+	struct ulptx_sge_pair buf[MAX_SKB_FRAGS / 2 + 1];
+
+	len = skb_headlen(skb) - start;
+	if (likely(len)) {
+		sgl->len0 = htonl(len);
+		sgl->addr0 = cpu_to_be64(addr[0] + start);
+		nfrags++;
+	} else {
+		sgl->len0 = htonl(si->frags[0].size);
+		sgl->addr0 = cpu_to_be64(addr[1]);
+	}
+
+	sgl->cmd_nsge = htonl(ULPTX_CMD(ULP_TX_SC_DSGL) | ULPTX_NSGE(nfrags));
+	if (likely(--nfrags == 0))
+		return;
+	/*
+	 * Most of the complexity below deals with the possibility we hit the
+	 * end of the queue in the middle of writing the SGL.  For this case
+	 * only we create the SGL in a temporary buffer and then copy it.
+	 */
+	to = (u8 *)end > (u8 *)q->stat ? buf : sgl->sge;
+
+	for (i = (nfrags != si->nr_frags); nfrags >= 2; nfrags -= 2, to++) {
+		to->len[0] = cpu_to_be32(si->frags[i].size);
+		to->len[1] = cpu_to_be32(si->frags[++i].size);
+		to->addr[0] = cpu_to_be64(addr[i]);
+		to->addr[1] = cpu_to_be64(addr[++i]);
+	}
+	if (nfrags) {
+		to->len[0] = cpu_to_be32(si->frags[i].size);
+		to->len[1] = cpu_to_be32(0);
+		to->addr[0] = cpu_to_be64(addr[i + 1]);
+	}
+	if (unlikely((u8 *)end > (u8 *)q->stat)) {
+		unsigned int part0 = (u8 *)q->stat - (u8 *)sgl->sge, part1;
+
+		if (likely(part0))
+			memcpy(sgl->sge, buf, part0);
+		part1 = (u8 *)end - (u8 *)q->stat;
+		memcpy(q->desc, (u8 *)buf + part0, part1);
+		end = (void *)q->desc + part1;
+	}
+	if ((uintptr_t)end & 8)           /* 0-pad to multiple of 16 */
+		*(u64 *)end = 0;
+}
+
+/**
+ *	ring_tx_db - check and potentially ring a Tx queue's doorbell
+ *	@adap: the adapter
+ *	@q: the Tx queue
+ *	@n: number of new descriptors to give to HW
+ *
+ *	Ring the doorbel for a Tx queue.
+ */
+static inline void ring_tx_db(struct adapter *adap, struct sge_txq *q, int n)
+{
+	wmb();            /* write descriptors before telling HW */
+	t4_write_reg(adap, MYPF_REG(SGE_PF_KDOORBELL),
+		     QID(q->cntxt_id) | PIDX(n));
+}
+
+/**
+ * 	inline_tx_skb - inline a packet's data into Tx descriptors
+ * 	@skb: the packet
+ * 	@q: the Tx queue where the packet will be inlined
+ * 	@pos: starting position in the Tx queue where to inline the packet
+ *
+ *	Inline a packet's contents directly into Tx descriptors, starting at
+ *	the given position within the Tx DMA ring.
+ *	Most of the complexity of this operation is dealing with wrap arounds
+ *	in the middle of the packet we want to inline.
+ */
+static void inline_tx_skb(const struct sk_buff *skb, const struct sge_txq *q,
+			  void *pos)
+{
+	u64 *p;
+	int left = (void *)q->stat - pos;
+
+	if (likely(skb->len <= left)) {
+		if (likely(!skb->data_len))
+			skb_copy_from_linear_data(skb, pos, skb->len);
+		else
+			skb_copy_bits(skb, 0, pos, skb->len);
+		pos += skb->len;
+	} else {
+		skb_copy_bits(skb, 0, pos, left);
+		skb_copy_bits(skb, left, q->desc, skb->len - left);
+		pos = (void *)q->desc + (skb->len - left);
+	}
+
+	/* 0-pad to multiple of 16 */
+	p = PTR_ALIGN(pos, 8);
+	if ((uintptr_t)p & 8)
+		*p = 0;
+}
+
+/*
+ * Figure out what HW csum a packet wants and return the appropriate control
+ * bits.
+ */
+static u64 hwcsum(const struct sk_buff *skb)
+{
+	int csum_type;
+	const struct iphdr *iph = ip_hdr(skb);
+
+	if (iph->version == 4) {
+		if (iph->protocol == IPPROTO_TCP)
+			csum_type = TX_CSUM_TCPIP;
+		else if (iph->protocol == IPPROTO_UDP)
+			csum_type = TX_CSUM_UDPIP;
+		else {
+nocsum:			/*
+			 * unknown protocol, disable HW csum
+			 * and hope a bad packet is detected
+			 */
+			return TXPKT_L4CSUM_DIS;
+		}
+	} else {
+		/*
+		 * this doesn't work with extension headers
+		 */
+		const struct ipv6hdr *ip6h = (const struct ipv6hdr *)iph;
+
+		if (ip6h->nexthdr == IPPROTO_TCP)
+			csum_type = TX_CSUM_TCPIP6;
+		else if (ip6h->nexthdr == IPPROTO_UDP)
+			csum_type = TX_CSUM_UDPIP6;
+		else
+			goto nocsum;
+	}
+
+	if (likely(csum_type >= TX_CSUM_TCPIP))
+		return TXPKT_CSUM_TYPE(csum_type) |
+			TXPKT_IPHDR_LEN(skb_network_header_len(skb)) |
+			TXPKT_ETHHDR_LEN(skb_network_offset(skb) - ETH_HLEN);
+	else {
+		int start = skb_transport_offset(skb);
+
+		return TXPKT_CSUM_TYPE(csum_type) | TXPKT_CSUM_START(start) |
+			TXPKT_CSUM_LOC(start + skb->csum_offset);
+	}
+}
+
+static void eth_txq_stop(struct sge_eth_txq *q)
+{
+	netif_tx_stop_queue(q->txq);
+	q->q.stops++;
+}
+
+static inline void txq_advance(struct sge_txq *q, unsigned int n)
+{
+	q->in_use += n;
+	q->pidx += n;
+	if (q->pidx >= q->size)
+		q->pidx -= q->size;
+}
+
+/**
+ *	t4_eth_xmit - add a packet to an Ethernet Tx queue
+ *	@skb: the packet
+ *	@dev: the egress net device
+ *
+ *	Add a packet to an SGE Ethernet Tx queue.  Runs with softirqs disabled.
+ */
+netdev_tx_t t4_eth_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+	u32 wr_mid;
+	u64 cntrl, *end;
+	int qidx, credits;
+	unsigned int flits, ndesc;
+	struct adapter *adap;
+	struct sge_eth_txq *q;
+	const struct port_info *pi;
+	struct fw_eth_tx_pkt_wr *wr;
+	struct cpl_tx_pkt_core *cpl;
+	const struct skb_shared_info *ssi;
+	dma_addr_t addr[MAX_SKB_FRAGS + 1];
+
+	/*
+	 * The chip min packet length is 10 octets but play safe and reject
+	 * anything shorter than an Ethernet header.
+	 */
+	if (unlikely(skb->len < ETH_HLEN)) {
+out_free:	dev_kfree_skb(skb);
+		return NETDEV_TX_OK;
+	}
+
+	pi = netdev_priv(dev);
+	adap = pi->adapter;
+	qidx = skb_get_queue_mapping(skb);
+	q = &adap->sge.ethtxq[qidx + pi->first_qset];
+
+	reclaim_completed_tx(adap, &q->q, true);
+
+	flits = calc_tx_flits(skb);
+	ndesc = flits_to_desc(flits);
+	credits = txq_avail(&q->q) - ndesc;
+
+	if (unlikely(credits < 0)) {
+		eth_txq_stop(q);
+		dev_err(adap->pdev_dev,
+			"%s: Tx ring %u full while queue awake!\n",
+			dev->name, qidx);
+		return NETDEV_TX_BUSY;
+	}
+
+	if (!is_eth_imm(skb) &&
+	    unlikely(map_skb(adap->pdev_dev, skb, addr) < 0)) {
+		q->mapping_err++;
+		goto out_free;
+	}
+
+	wr_mid = FW_WR_LEN16(DIV_ROUND_UP(flits, 2));
+	if (unlikely(credits < ETHTXQ_STOP_THRES)) {
+		eth_txq_stop(q);
+		wr_mid |= FW_WR_EQUEQ | FW_WR_EQUIQ;
+	}
+
+	wr = (void *)&q->q.desc[q->q.pidx];
+	wr->equiq_to_len16 = htonl(wr_mid);
+	wr->r3 = cpu_to_be64(0);
+	end = (u64 *)wr + flits;
+
+	ssi = skb_shinfo(skb);
+	if (ssi->gso_size) {
+		struct cpl_tx_pkt_lso *lso = (void *)wr;
+		bool v6 = (ssi->gso_type & SKB_GSO_TCPV6) != 0;
+		int l3hdr_len = skb_network_header_len(skb);
+		int eth_xtra_len = skb_network_offset(skb) - ETH_HLEN;
+
+		wr->op_immdlen = htonl(FW_WR_OP(FW_ETH_TX_PKT_WR) |
+				       FW_WR_IMMDLEN(sizeof(*lso)));
+		lso->lso_ctrl = htonl(LSO_OPCODE(CPL_TX_PKT_LSO) |
+				      LSO_FIRST_SLICE | LSO_LAST_SLICE |
+				      LSO_IPV6(v6) |
+				      LSO_ETHHDR_LEN(eth_xtra_len / 4) |
+				      LSO_IPHDR_LEN(l3hdr_len / 4) |
+				      LSO_TCPHDR_LEN(tcp_hdr(skb)->doff));
+		lso->ipid_ofst = htons(0);
+		lso->mss = htons(ssi->gso_size);
+		lso->seqno_offset = htonl(0);
+		lso->len = htonl(skb->len);
+		cpl = (void *)(lso + 1);
+		cntrl = TXPKT_CSUM_TYPE(v6 ? TX_CSUM_TCPIP6 : TX_CSUM_TCPIP) |
+			TXPKT_IPHDR_LEN(l3hdr_len) |
+			TXPKT_ETHHDR_LEN(eth_xtra_len);
+		q->tso++;
+		q->tx_cso += ssi->gso_segs;
+	} else {
+		int len;
+
+		len = is_eth_imm(skb) ? skb->len + sizeof(*cpl) : sizeof(*cpl);
+		wr->op_immdlen = htonl(FW_WR_OP(FW_ETH_TX_PKT_WR) |
+				       FW_WR_IMMDLEN(len));
+		cpl = (void *)(wr + 1);
+		if (skb->ip_summed == CHECKSUM_PARTIAL) {
+			cntrl = hwcsum(skb) | TXPKT_IPCSUM_DIS;
+			q->tx_cso++;
+		} else
+			cntrl = TXPKT_L4CSUM_DIS | TXPKT_IPCSUM_DIS;
+	}
+
+	if (vlan_tx_tag_present(skb)) {
+		q->vlan_ins++;
+		cntrl |= TXPKT_VLAN_VLD | TXPKT_VLAN(vlan_tx_tag_get(skb));
+	}
+
+	cpl->ctrl0 = htonl(TXPKT_OPCODE(CPL_TX_PKT_XT) |
+			   TXPKT_INTF(pi->tx_chan) | TXPKT_PF(0));
+	cpl->pack = htons(0);
+	cpl->len = htons(skb->len);
+	cpl->ctrl1 = cpu_to_be64(cntrl);
+
+	if (is_eth_imm(skb)) {
+		inline_tx_skb(skb, &q->q, cpl + 1);
+		dev_kfree_skb(skb);
+	} else {
+		int last_desc;
+
+		write_sgl(skb, &q->q, (struct ulptx_sgl *)(cpl + 1), end, 0,
+			  addr);
+		skb_orphan(skb);
+
+		last_desc = q->q.pidx + ndesc - 1;
+		if (last_desc >= q->q.size)
+			last_desc -= q->q.size;
+		q->q.sdesc[last_desc].skb = skb;
+		q->q.sdesc[last_desc].sgl = (struct ulptx_sgl *)(cpl + 1);
+	}
+
+	txq_advance(&q->q, ndesc);
+
+	ring_tx_db(adap, &q->q, ndesc);
+	return NETDEV_TX_OK;
+}
+
+/**
+ *	reclaim_completed_tx_imm - reclaim completed control-queue Tx descs
+ *	@q: the SGE control Tx queue
+ *
+ *	This is a variant of reclaim_completed_tx() that is used for Tx queues
+ *	that send only immediate data (presently just the control queues) and
+ *	thus do not have any sk_buffs to release.
+ */
+static inline void reclaim_completed_tx_imm(struct sge_txq *q)
+{
+	int hw_cidx = ntohs(q->stat->cidx);
+	int reclaim = hw_cidx - q->cidx;
+
+	if (reclaim < 0)
+		reclaim += q->size;
+
+	q->in_use -= reclaim;
+	q->cidx = hw_cidx;
+}
+
+/**
+ *	is_imm - check whether a packet can be sent as immediate data
+ *	@skb: the packet
+ *
+ *	Returns true if a packet can be sent as a WR with immediate data.
+ */
+static inline int is_imm(const struct sk_buff *skb)
+{
+	return skb->len <= MAX_CTRL_WR_LEN;
+}
+
+/**
+ *	ctrlq_check_stop - check if a control queue is full and should stop
+ *	@q: the queue
+ *	@wr: most recent WR written to the queue
+ *
+ *	Check if a control queue has become full and should be stopped.
+ *	We clean up control queue descriptors very lazily, only when we are out.
+ *	If the queue is still full after reclaiming any completed descriptors
+ *	we suspend it and have the last WR wake it up.
+ */
+static void ctrlq_check_stop(struct sge_ctrl_txq *q, struct fw_wr_hdr *wr)
+{
+	reclaim_completed_tx_imm(&q->q);
+	if (unlikely(txq_avail(&q->q) < TXQ_STOP_THRES)) {
+		wr->lo |= htonl(FW_WR_EQUEQ | FW_WR_EQUIQ);
+		q->q.stops++;
+		q->full = 1;
+	}
+}
+
+/**
+ *	ctrl_xmit - send a packet through an SGE control Tx queue
+ *	@q: the control queue
+ *	@skb: the packet
+ *
+ *	Send a packet through an SGE control Tx queue.  Packets sent through
+ *	a control queue must fit entirely as immediate data.
+ */
+static int ctrl_xmit(struct sge_ctrl_txq *q, struct sk_buff *skb)
+{
+	unsigned int ndesc;
+	struct fw_wr_hdr *wr;
+
+	if (unlikely(!is_imm(skb))) {
+		WARN_ON(1);
+		dev_kfree_skb(skb);
+		return NET_XMIT_DROP;
+	}
+
+	ndesc = DIV_ROUND_UP(skb->len, sizeof(struct tx_desc));
+	spin_lock(&q->sendq.lock);
+
+	if (unlikely(q->full)) {
+		skb->priority = ndesc;                  /* save for restart */
+		__skb_queue_tail(&q->sendq, skb);
+		spin_unlock(&q->sendq.lock);
+		return NET_XMIT_CN;
+	}
+
+	wr = (struct fw_wr_hdr *)&q->q.desc[q->q.pidx];
+	inline_tx_skb(skb, &q->q, wr);
+
+	txq_advance(&q->q, ndesc);
+	if (unlikely(txq_avail(&q->q) < TXQ_STOP_THRES))
+		ctrlq_check_stop(q, wr);
+
+	ring_tx_db(q->adap, &q->q, ndesc);
+	spin_unlock(&q->sendq.lock);
+
+	kfree_skb(skb);
+	return NET_XMIT_SUCCESS;
+}
+
+/**
+ *	restart_ctrlq - restart a suspended control queue
+ *	@data: the control queue to restart
+ *
+ *	Resumes transmission on a suspended Tx control queue.
+ */
+static void restart_ctrlq(unsigned long data)
+{
+	struct sk_buff *skb;
+	unsigned int written = 0;
+	struct sge_ctrl_txq *q = (struct sge_ctrl_txq *)data;
+
+	spin_lock(&q->sendq.lock);
+	reclaim_completed_tx_imm(&q->q);
+	BUG_ON(txq_avail(&q->q) < TXQ_STOP_THRES);  /* q should be empty */
+
+	while ((skb = __skb_dequeue(&q->sendq)) != NULL) {
+		struct fw_wr_hdr *wr;
+		unsigned int ndesc = skb->priority;     /* previously saved */
+
+		/*
+		 * Write descriptors and free skbs outside the lock to limit
+		 * wait times.  q->full is still set so new skbs will be queued.
+		 */
+		spin_unlock(&q->sendq.lock);
+
+		wr = (struct fw_wr_hdr *)&q->q.desc[q->q.pidx];
+		inline_tx_skb(skb, &q->q, wr);
+		kfree_skb(skb);
+
+		written += ndesc;
+		txq_advance(&q->q, ndesc);
+		if (unlikely(txq_avail(&q->q) < TXQ_STOP_THRES)) {
+			unsigned long old = q->q.stops;
+
+			ctrlq_check_stop(q, wr);
+			if (q->q.stops != old) {          /* suspended anew */
+				spin_lock(&q->sendq.lock);
+				goto ringdb;
+			}
+		}
+		if (written > 16) {
+			ring_tx_db(q->adap, &q->q, written);
+			written = 0;
+		}
+		spin_lock(&q->sendq.lock);
+	}
+	q->full = 0;
+ringdb: if (written)
+		ring_tx_db(q->adap, &q->q, written);
+	spin_unlock(&q->sendq.lock);
+}
+
+/**
+ *	t4_mgmt_tx - send a management message
+ *	@adap: the adapter
+ *	@skb: the packet containing the management message
+ *
+ *	Send a management message through control queue 0.
+ */
+int t4_mgmt_tx(struct adapter *adap, struct sk_buff *skb)
+{
+	int ret;
+
+	local_bh_disable();
+	ret = ctrl_xmit(&adap->sge.ctrlq[0], skb);
+	local_bh_enable();
+	return ret;
+}
+
+/**
+ *	deferred_unmap_destructor - unmap a packet when it is freed
+ *	@skb: the packet
+ *
+ *	This is the packet destructor used for Tx packets that need to remain
+ *	mapped until they are freed rather than until their Tx descriptors are
+ *	freed.
+ */
+static void deferred_unmap_destructor(struct sk_buff *skb)
+{
+	unmap_skb(skb->dev->dev.parent, skb, (dma_addr_t *)skb->head);
+}
+
+/**
+ *	is_ofld_imm - check whether a packet can be sent as immediate data
+ *	@skb: the packet
+ *
+ *	Returns true if a packet can be sent as an offload WR with immediate
+ *	data.  We currently use the same limit as for Ethernet packets.
+ */
+static inline int is_ofld_imm(const struct sk_buff *skb)
+{
+	return skb->len <= MAX_IMM_TX_PKT_LEN;
+}
+
+/**
+ *	calc_tx_flits_ofld - calculate # of flits for an offload packet
+ *	@skb: the packet
+ *
+ * 	Returns the number of flits needed for the given offload packet.
+ * 	These packets are already fully constructed and no additional headers
+ * 	will be added.
+ */
+static inline unsigned int calc_tx_flits_ofld(const struct sk_buff *skb)
+{
+	unsigned int flits, cnt;
+
+	if (is_ofld_imm(skb))
+		return DIV_ROUND_UP(skb->len, 8);
+
+	flits = skb_transport_offset(skb) / 8U;   /* headers */
+	cnt = skb_shinfo(skb)->nr_frags;
+	if (skb->tail != skb->transport_header)
+		cnt++;
+	return flits + sgl_len(cnt);
+}
+
+/**
+ *	txq_stop_maperr - stop a Tx queue due to I/O MMU exhaustion
+ *	@adap: the adapter
+ *	@q: the queue to stop
+ *
+ *	Mark a Tx queue stopped due to I/O MMU exhaustion and resulting
+ *	inability to map packets.  A periodic timer attempts to restart
+ *	queues so marked.
+ */
+static void txq_stop_maperr(struct sge_ofld_txq *q)
+{
+	q->mapping_err++;
+	q->q.stops++;
+	set_bit(q->q.cntxt_id, q->adap->sge.txq_maperr);
+}
+
+/**
+ *	ofldtxq_stop - stop an offload Tx queue that has become full
+ *	@q: the queue to stop
+ *	@skb: the packet causing the queue to become full
+ *
+ *	Stops an offload Tx queue that has become full and modifies the packet
+ *	being written to request a wakeup.
+ */
+static void ofldtxq_stop(struct sge_ofld_txq *q, struct sk_buff *skb)
+{
+	struct fw_wr_hdr *wr = (struct fw_wr_hdr *)skb->data;
+
+	wr->lo |= htonl(FW_WR_EQUEQ | FW_WR_EQUIQ);
+	q->q.stops++;
+	q->full = 1;
+}
+
+/**
+ *	service_ofldq - restart a suspended offload queue
+ *	@q: the offload queue
+ *
+ *	Services an offload Tx queue by moving packets from its packet queue
+ *	to the HW Tx ring.  The function starts and ends with the queue locked.
+ */
+static void service_ofldq(struct sge_ofld_txq *q)
+{
+	u64 *pos;
+	int credits;
+	struct sk_buff *skb;
+	unsigned int written = 0;
+	unsigned int flits, ndesc;
+
+	while ((skb = skb_peek(&q->sendq)) != NULL && !q->full) {
+		/*
+		 * We drop the lock but leave skb on sendq, thus retaining
+		 * exclusive access to the state of the queue.
+		 */
+		spin_unlock(&q->sendq.lock);
+
+		reclaim_completed_tx(q->adap, &q->q, false);
+
+		flits = skb->priority;                /* previously saved */
+		ndesc = flits_to_desc(flits);
+		credits = txq_avail(&q->q) - ndesc;
+		BUG_ON(credits < 0);
+		if (unlikely(credits < TXQ_STOP_THRES))
+			ofldtxq_stop(q, skb);
+
+		pos = (u64 *)&q->q.desc[q->q.pidx];
+		if (is_ofld_imm(skb))
+			inline_tx_skb(skb, &q->q, pos);
+		else if (map_skb(q->adap->pdev_dev, skb,
+				 (dma_addr_t *)skb->head)) {
+			txq_stop_maperr(q);
+			spin_lock(&q->sendq.lock);
+			break;
+		} else {
+			int last_desc, hdr_len = skb_transport_offset(skb);
+
+			memcpy(pos, skb->data, hdr_len);
+			write_sgl(skb, &q->q, (void *)pos + hdr_len,
+				  pos + flits, hdr_len,
+				  (dma_addr_t *)skb->head);
+
+			if (need_skb_unmap()) {
+				skb->dev = q->adap->port[0];
+				skb->destructor = deferred_unmap_destructor;
+			}
+
+			last_desc = q->q.pidx + ndesc - 1;
+			if (last_desc >= q->q.size)
+				last_desc -= q->q.size;
+			q->q.sdesc[last_desc].skb = skb;
+		}
+
+		txq_advance(&q->q, ndesc);
+		written += ndesc;
+		if (unlikely(written > 32)) {
+			ring_tx_db(q->adap, &q->q, written);
+			written = 0;
+		}
+
+		spin_lock(&q->sendq.lock);
+		__skb_unlink(skb, &q->sendq);
+		if (is_ofld_imm(skb))
+			kfree_skb(skb);
+	}
+	if (likely(written))
+		ring_tx_db(q->adap, &q->q, written);
+}
+
+/**
+ *	ofld_xmit - send a packet through an offload queue
+ *	@q: the Tx offload queue
+ *	@skb: the packet
+ *
+ *	Send an offload packet through an SGE offload queue.
+ */
+static int ofld_xmit(struct sge_ofld_txq *q, struct sk_buff *skb)
+{
+	skb->priority = calc_tx_flits_ofld(skb);       /* save for restart */
+	spin_lock(&q->sendq.lock);
+	__skb_queue_tail(&q->sendq, skb);
+	if (q->sendq.qlen == 1)
+		service_ofldq(q);
+	spin_unlock(&q->sendq.lock);
+	return NET_XMIT_SUCCESS;
+}
+
+/**
+ *	restart_ofldq - restart a suspended offload queue
+ *	@data: the offload queue to restart
+ *
+ *	Resumes transmission on a suspended Tx offload queue.
+ */
+static void restart_ofldq(unsigned long data)
+{
+	struct sge_ofld_txq *q = (struct sge_ofld_txq *)data;
+
+	spin_lock(&q->sendq.lock);
+	q->full = 0;            /* the queue actually is completely empty now */
+	service_ofldq(q);
+	spin_unlock(&q->sendq.lock);
+}
+
+/**
+ *	skb_txq - return the Tx queue an offload packet should use
+ *	@skb: the packet
+ *
+ *	Returns the Tx queue an offload packet should use as indicated by bits
+ *	1-15 in the packet's queue_mapping.
+ */
+static inline unsigned int skb_txq(const struct sk_buff *skb)
+{
+	return skb->queue_mapping >> 1;
+}
+
+/**
+ *	is_ctrl_pkt - return whether an offload packet is a control packet
+ *	@skb: the packet
+ *
+ *	Returns whether an offload packet should use an OFLD or a CTRL
+ *	Tx queue as indicated by bit 0 in the packet's queue_mapping.
+ */
+static inline unsigned int is_ctrl_pkt(const struct sk_buff *skb)
+{
+	return skb->queue_mapping & 1;
+}
+
+static inline int ofld_send(struct adapter *adap, struct sk_buff *skb)
+{
+	unsigned int idx = skb_txq(skb);
+
+	if (unlikely(is_ctrl_pkt(skb)))
+		return ctrl_xmit(&adap->sge.ctrlq[idx], skb);
+	return ofld_xmit(&adap->sge.ofldtxq[idx], skb);
+}
+
+/**
+ *	t4_ofld_send - send an offload packet
+ *	@adap: the adapter
+ *	@skb: the packet
+ *
+ *	Sends an offload packet.  We use the packet queue_mapping to select the
+ *	appropriate Tx queue as follows: bit 0 indicates whether the packet
+ *	should be sent as regular or control, bits 1-15 select the queue.
+ */
+int t4_ofld_send(struct adapter *adap, struct sk_buff *skb)
+{
+	int ret;
+
+	local_bh_disable();
+	ret = ofld_send(adap, skb);
+	local_bh_enable();
+	return ret;
+}
+
+/**
+ *	cxgb4_ofld_send - send an offload packet
+ *	@dev: the net device
+ *	@skb: the packet
+ *
+ *	Sends an offload packet.  This is an exported version of @t4_ofld_send,
+ *	intended for ULDs.
+ */
+int cxgb4_ofld_send(struct net_device *dev, struct sk_buff *skb)
+{
+	return t4_ofld_send(netdev2adap(dev), skb);
+}
+EXPORT_SYMBOL(cxgb4_ofld_send);
+
+static inline void copy_frags(struct skb_shared_info *ssi,
+			      const struct pkt_gl *gl, unsigned int offset)
+{
+	unsigned int n;
+
+	/* usually there's just one frag */
+	ssi->frags[0].page = gl->frags[0].page;
+	ssi->frags[0].page_offset = gl->frags[0].page_offset + offset;
+	ssi->frags[0].size = gl->frags[0].size - offset;
+	ssi->nr_frags = gl->nfrags;
+	n = gl->nfrags - 1;
+	if (n)
+		memcpy(&ssi->frags[1], &gl->frags[1], n * sizeof(skb_frag_t));
+
+	/* get a reference to the last page, we don't own it */
+	get_page(gl->frags[n].page);
+}
+
+/**
+ *	cxgb4_pktgl_to_skb - build an sk_buff from a packet gather list
+ *	@gl: the gather list
+ *	@skb_len: size of sk_buff main body if it carries fragments
+ *	@pull_len: amount of data to move to the sk_buff's main body
+ *
+ *	Builds an sk_buff from the given packet gather list.  Returns the
+ *	sk_buff or %NULL if sk_buff allocation failed.
+ */
+struct sk_buff *cxgb4_pktgl_to_skb(const struct pkt_gl *gl,
+				   unsigned int skb_len, unsigned int pull_len)
+{
+	struct sk_buff *skb;
+
+	/*
+	 * Below we rely on RX_COPY_THRES being less than the smallest Rx buffer
+	 * size, which is expected since buffers are at least PAGE_SIZEd.
+	 * In this case packets up to RX_COPY_THRES have only one fragment.
+	 */
+	if (gl->tot_len <= RX_COPY_THRES) {
+		skb = dev_alloc_skb(gl->tot_len);
+		if (unlikely(!skb))
+			goto out;
+		__skb_put(skb, gl->tot_len);
+		skb_copy_to_linear_data(skb, gl->va, gl->tot_len);
+	} else {
+		skb = dev_alloc_skb(skb_len);
+		if (unlikely(!skb))
+			goto out;
+		__skb_put(skb, pull_len);
+		skb_copy_to_linear_data(skb, gl->va, pull_len);
+
+		copy_frags(skb_shinfo(skb), gl, pull_len);
+		skb->len = gl->tot_len;
+		skb->data_len = skb->len - pull_len;
+		skb->truesize += skb->data_len;
+	}
+out:	return skb;
+}
+EXPORT_SYMBOL(cxgb4_pktgl_to_skb);
+
+/**
+ *	t4_pktgl_free - free a packet gather list
+ *	@gl: the gather list
+ *
+ *	Releases the pages of a packet gather list.  We do not own the last
+ *	page on the list and do not free it.
+ */
+void t4_pktgl_free(const struct pkt_gl *gl)
+{
+	int n;
+	const skb_frag_t *p;
+
+	for (p = gl->frags, n = gl->nfrags - 1; n--; p++)
+		put_page(p->page);
+}
+
+/*
+ * Process an MPS trace packet.  Give it an unused protocol number so it won't
+ * be delivered to anyone and send it to the stack for capture.
+ */
+static noinline int handle_trace_pkt(struct adapter *adap,
+				     const struct pkt_gl *gl)
+{
+	struct sk_buff *skb;
+	struct cpl_trace_pkt *p;
+
+	skb = cxgb4_pktgl_to_skb(gl, RX_PULL_LEN, RX_PULL_LEN);
+	if (unlikely(!skb)) {
+		t4_pktgl_free(gl);
+		return 0;
+	}
+
+	p = (struct cpl_trace_pkt *)skb->data;
+	__skb_pull(skb, sizeof(*p));
+	skb_reset_mac_header(skb);
+	skb->protocol = htons(0xffff);
+	skb->dev = adap->port[0];
+	netif_receive_skb(skb);
+	return 0;
+}
+
+static void do_gro(struct sge_eth_rxq *rxq, const struct pkt_gl *gl,
+		   const struct cpl_rx_pkt *pkt)
+{
+	int ret;
+	struct sk_buff *skb;
+
+	skb = napi_get_frags(&rxq->rspq.napi);
+	if (unlikely(!skb)) {
+		t4_pktgl_free(gl);
+		rxq->stats.rx_drops++;
+		return;
+	}
+
+	copy_frags(skb_shinfo(skb), gl, RX_PKT_PAD);
+	skb->len = gl->tot_len - RX_PKT_PAD;
+	skb->data_len = skb->len;
+	skb->truesize += skb->data_len;
+	skb->ip_summed = CHECKSUM_UNNECESSARY;
+	skb_record_rx_queue(skb, rxq->rspq.idx);
+
+	if (unlikely(pkt->vlan_ex)) {
+		struct port_info *pi = netdev_priv(rxq->rspq.netdev);
+		struct vlan_group *grp = pi->vlan_grp;
+
+		rxq->stats.vlan_ex++;
+		if (likely(grp)) {
+			ret = vlan_gro_frags(&rxq->rspq.napi, grp,
+					     ntohs(pkt->vlan));
+			goto stats;
+		}
+	}
+	ret = napi_gro_frags(&rxq->rspq.napi);
+stats:	if (ret == GRO_HELD)
+		rxq->stats.lro_pkts++;
+	else if (ret == GRO_MERGED || ret == GRO_MERGED_FREE)
+		rxq->stats.lro_merged++;
+	rxq->stats.pkts++;
+	rxq->stats.rx_cso++;
+}
+
+/**
+ *	t4_ethrx_handler - process an ingress ethernet packet
+ *	@q: the response queue that received the packet
+ *	@rsp: the response queue descriptor holding the RX_PKT message
+ *	@si: the gather list of packet fragments
+ *
+ *	Process an ingress ethernet packet and deliver it to the stack.
+ */
+int t4_ethrx_handler(struct sge_rspq *q, const __be64 *rsp,
+		     const struct pkt_gl *si)
+{
+	bool csum_ok;
+	struct sk_buff *skb;
+	struct port_info *pi;
+	const struct cpl_rx_pkt *pkt;
+	struct sge_eth_rxq *rxq = container_of(q, struct sge_eth_rxq, rspq);
+
+	if (unlikely(*(u8 *)rsp == CPL_TRACE_PKT))
+		return handle_trace_pkt(q->adap, si);
+
+	pkt = (void *)&rsp[1];
+	csum_ok = pkt->csum_calc && !pkt->err_vec;
+	if ((pkt->l2info & htonl(RXF_TCP)) &&
+	    (q->netdev->features & NETIF_F_GRO) && csum_ok && !pkt->ip_frag) {
+		do_gro(rxq, si, pkt);
+		return 0;
+	}
+
+	if (si->tot_len <= RX_COPY_THRES) {
+		/* small packets have only one fragment */
+		skb = dev_alloc_skb(si->frags[0].size);
+		if (!skb)
+			goto nomem;
+		__skb_put(skb, si->frags[0].size);
+		skb_copy_to_linear_data(skb, si->va, si->frags[0].size);
+	} else {
+		skb = dev_alloc_skb(RX_PKT_SKB_LEN);
+		if (!skb)
+			goto nomem;
+		__skb_put(skb, RX_PULL_LEN);
+		skb_copy_to_linear_data(skb, si->va, RX_PULL_LEN);
+
+		copy_frags(skb_shinfo(skb), si, RX_PULL_LEN);
+		skb->len = si->tot_len;
+		skb->data_len = skb->len - RX_PULL_LEN;
+		skb->truesize += skb->data_len;
+	}
+
+	__skb_pull(skb, RX_PKT_PAD);      /* remove ethernet header padding */
+	skb->protocol = eth_type_trans(skb, q->netdev);
+	skb_record_rx_queue(skb, q->idx);
+	pi = netdev_priv(skb->dev);
+	rxq->stats.pkts++;
+
+	if (csum_ok && (pi->rx_offload & RX_CSO) &&
+	    (pkt->l2info & htonl(RXF_UDP | RXF_TCP))) {
+		if (!pkt->ip_frag)
+			skb->ip_summed = CHECKSUM_UNNECESSARY;
+		else {
+			__sum16 c = (__force __sum16)pkt->csum;
+			skb->csum = csum_unfold(c);
+			skb->ip_summed = CHECKSUM_COMPLETE;
+		}
+		rxq->stats.rx_cso++;
+	} else
+		skb->ip_summed = CHECKSUM_NONE;
+
+	if (unlikely(pkt->vlan_ex)) {
+		struct vlan_group *grp = pi->vlan_grp;
+
+		rxq->stats.vlan_ex++;
+		if (likely(grp))
+			vlan_hwaccel_receive_skb(skb, grp, ntohs(pkt->vlan));
+		else
+			dev_kfree_skb_any(skb);
+	} else
+		netif_receive_skb(skb);
+
+	return 0;
+
+nomem:	t4_pktgl_free(si);
+	rxq->stats.rx_drops++;
+	return 0;
+}
+
+/**
+ *	restore_rx_bufs - put back a packet's Rx buffers
+ *	@si: the packet gather list
+ *	@q: the SGE free list
+ *	@frags: number of FL buffers to restore
+ *
+ *	Puts back on an FL the Rx buffers associated with @si.  The buffers
+ *	have already been unmapped and are left unmapped, we mark them so to
+ *	prevent further unmapping attempts.
+ *
+ *	This function undoes a series of @unmap_rx_buf calls when we find out
+ *	that the current packet can't be processed right away afterall and we
+ *	need to come back to it later.  This is a very rare event and there's
+ *	no effort to make this particularly efficient.
+ */
+static void restore_rx_bufs(const struct pkt_gl *si, struct sge_fl *q,
+			    int frags)
+{
+	struct rx_sw_desc *d;
+
+	while (frags--) {
+		if (q->cidx == 0)
+			q->cidx = q->size - 1;
+		else
+			q->cidx--;
+		d = &q->sdesc[q->cidx];
+		d->page = si->frags[frags].page;
+		d->dma_addr |= RX_UNMAPPED_BUF;
+		q->avail++;
+	}
+}
+
+/**
+ *	is_new_response - check if a response is newly written
+ *	@r: the response descriptor
+ *	@q: the response queue
+ *
+ *	Returns true if a response descriptor contains a yet unprocessed
+ *	response.
+ */
+static inline bool is_new_response(const struct rsp_ctrl *r,
+				   const struct sge_rspq *q)
+{
+	return RSPD_GEN(r->type_gen) == q->gen;
+}
+
+/**
+ *	rspq_next - advance to the next entry in a response queue
+ *	@q: the queue
+ *
+ *	Updates the state of a response queue to advance it to the next entry.
+ */
+static inline void rspq_next(struct sge_rspq *q)
+{
+	q->cur_desc = (void *)q->cur_desc + q->iqe_len;
+	if (unlikely(++q->cidx == q->size)) {
+		q->cidx = 0;
+		q->gen ^= 1;
+		q->cur_desc = q->desc;
+	}
+}
+
+/**
+ *	process_responses - process responses from an SGE response queue
+ *	@q: the ingress queue to process
+ *	@budget: how many responses can be processed in this round
+ *
+ *	Process responses from an SGE response queue up to the supplied budget.
+ *	Responses include received packets as well as control messages from FW
+ *	or HW.
+ *
+ *	Additionally choose the interrupt holdoff time for the next interrupt
+ *	on this queue.  If the system is under memory shortage use a fairly
+ *	long delay to help recovery.
+ */
+int process_responses(struct sge_rspq *q, int budget)
+{
+	int ret, rsp_type;
+	int budget_left = budget;
+	const struct rsp_ctrl *rc;
+	struct sge_eth_rxq *rxq = container_of(q, struct sge_eth_rxq, rspq);
+
+	while (likely(budget_left)) {
+		rc = (void *)q->cur_desc + (q->iqe_len - sizeof(*rc));
+		if (!is_new_response(rc, q))
+			break;
+
+		rmb();
+		rsp_type = RSPD_TYPE(rc->type_gen);
+		if (likely(rsp_type == RSP_TYPE_FLBUF)) {
+			skb_frag_t *fp;
+			struct pkt_gl si;
+			const struct rx_sw_desc *rsd;
+			u32 len = ntohl(rc->pldbuflen_qid), bufsz, frags;
+
+			if (len & RSPD_NEWBUF) {
+				if (likely(q->offset > 0)) {
+					free_rx_bufs(q->adap, &rxq->fl, 1);
+					q->offset = 0;
+				}
+				len &= RSPD_LEN;
+			}
+			si.tot_len = len;
+
+			/* gather packet fragments */
+			for (frags = 0, fp = si.frags; ; frags++, fp++) {
+				rsd = &rxq->fl.sdesc[rxq->fl.cidx];
+				bufsz = get_buf_size(rsd);
+				fp->page = rsd->page;
+				fp->page_offset = q->offset;
+				fp->size = min(bufsz, len);
+				len -= fp->size;
+				if (!len)
+					break;
+				unmap_rx_buf(q->adap, &rxq->fl);
+			}
+
+			/*
+			 * Last buffer remains mapped so explicitly make it
+			 * coherent for CPU access.
+			 */
+			dma_sync_single_for_cpu(q->adap->pdev_dev,
+						get_buf_addr(rsd),
+						fp->size, DMA_FROM_DEVICE);
+
+			si.va = page_address(si.frags[0].page) +
+				si.frags[0].page_offset;
+			prefetch(si.va);
+
+			si.nfrags = frags + 1;
+			ret = q->handler(q, q->cur_desc, &si);
+			if (likely(ret == 0))
+				q->offset += ALIGN(fp->size, FL_ALIGN);
+			else
+				restore_rx_bufs(&si, &rxq->fl, frags);
+		} else if (likely(rsp_type == RSP_TYPE_CPL)) {
+			ret = q->handler(q, q->cur_desc, NULL);
+		} else {
+			ret = q->handler(q, (const __be64 *)rc, CXGB4_MSG_AN);
+		}
+
+		if (unlikely(ret)) {
+			/* couldn't process descriptor, back off for recovery */
+			q->next_intr_params = QINTR_TIMER_IDX(NOMEM_TMR_IDX);
+			break;
+		}
+
+		rspq_next(q);
+		budget_left--;
+	}
+
+	if (q->offset >= 0 && rxq->fl.size - rxq->fl.avail >= 16)
+		replenish_fl(q->adap, &rxq->fl, GFP_ATOMIC);
+	return budget - budget_left;
+}
+
+/**
+ *	napi_rx_handler - the NAPI handler for Rx processing
+ *	@napi: the napi instance
+ *	@budget: how many packets we can process in this round
+ *
+ *	Handler for new data events when using NAPI.  This does not need any
+ *	locking or protection from interrupts as data interrupts are off at
+ *	this point and other adapter interrupts do not interfere (the latter
+ *	in not a concern at all with MSI-X as non-data interrupts then have
+ *	a separate handler).
+ */
+static int napi_rx_handler(struct napi_struct *napi, int budget)
+{
+	unsigned int params;
+	struct sge_rspq *q = container_of(napi, struct sge_rspq, napi);
+	int work_done = process_responses(q, budget);
+
+	if (likely(work_done < budget)) {
+		napi_complete(napi);
+		params = q->next_intr_params;
+		q->next_intr_params = q->intr_params;
+	} else
+		params = QINTR_TIMER_IDX(7);
+
+	t4_write_reg(q->adap, MYPF_REG(SGE_PF_GTS), CIDXINC(work_done) |
+		     INGRESSQID((u32)q->cntxt_id) | SEINTARM(params));
+	return work_done;
+}
+
+/*
+ * Returns true if a NAPI instance is already scheduled for polling.
+ */
+static inline int napi_is_scheduled(const struct napi_struct *napi)
+{
+	return test_bit(NAPI_STATE_SCHED, &napi->state);
+}
+
+/*
+ * The MSI-X interrupt handler for an SGE response queue.
+ */
+irqreturn_t t4_sge_intr_msix(int irq, void *cookie)
+{
+	struct sge_rspq *q = cookie;
+
+	spin_lock(&q->lock);
+	napi_schedule(&q->napi);
+	spin_unlock(&q->lock);
+	return IRQ_HANDLED;
+}
+
+/*
+ * Process the indirect interrupt entries in the interrupt queue and kick off
+ * NAPI for each queue that has generated an entry.
+ */
+static unsigned int process_intrq(struct adapter *adap)
+{
+	unsigned int credits;
+	const struct rsp_ctrl *rc;
+	struct sge_rspq *q = &adap->sge.intrq;
+
+	for (credits = 0; ; credits++) {
+		rc = (void *)q->cur_desc + (q->iqe_len - sizeof(*rc));
+		if (!is_new_response(rc, q))
+			break;
+
+		rmb();
+		if (RSPD_TYPE(rc->type_gen) == RSP_TYPE_INTR) {
+			unsigned int qid = ntohl(rc->pldbuflen_qid);
+			struct sge_rspq *srcq = adap->sge.ingr_map[qid];
+
+			spin_lock(&srcq->lock);
+			napi_schedule(&srcq->napi);
+			spin_unlock(&srcq->lock);
+		}
+
+		rspq_next(q);
+	}
+
+	t4_write_reg(adap, MYPF_REG(SGE_PF_GTS), CIDXINC(credits) |
+		     INGRESSQID(q->cntxt_id) | SEINTARM(q->intr_params));
+	return credits;
+}
+
+/*
+ * The MSI interrupt handler, which handles data events from SGE response queues
+ * as well as error and other async events as they all use the same MSI vector.
+ */
+static irqreturn_t t4_intr_msi(int irq, void *cookie)
+{
+	struct adapter *adap = cookie;
+
+	t4_slow_intr_handler(adap);
+	process_intrq(adap);
+	return IRQ_HANDLED;
+}
+
+/*
+ * Interrupt handler for legacy INTx interrupts.
+ * Handles data events from SGE response queues as well as error and other
+ * async events as they all use the same interrupt line.
+ */
+static irqreturn_t t4_intr_intx(int irq, void *cookie)
+{
+	struct adapter *adap = cookie;
+
+	t4_write_reg(adap, MYPF_REG(PCIE_PF_CLI), 0);
+	if (t4_slow_intr_handler(adap) | process_intrq(adap))
+		return IRQ_HANDLED;
+	return IRQ_NONE;             /* probably shared interrupt */
+}
+
+/**
+ *	t4_intr_handler - select the top-level interrupt handler
+ *	@adap: the adapter
+ *
+ *	Selects the top-level interrupt handler based on the type of interrupts
+ *	(MSI-X, MSI, or INTx).
+ */
+irq_handler_t t4_intr_handler(struct adapter *adap)
+{
+	if (adap->flags & USING_MSIX)
+		return t4_sge_intr_msix;
+	if (adap->flags & USING_MSI)
+		return t4_intr_msi;
+	return t4_intr_intx;
+}
+
+static void sge_rx_timer_cb(unsigned long data)
+{
+	unsigned long m;
+	unsigned int i, cnt[2];
+	struct adapter *adap = (struct adapter *)data;
+	struct sge *s = &adap->sge;
+
+	for (i = 0; i < ARRAY_SIZE(s->starving_fl); i++)
+		for (m = s->starving_fl[i]; m; m &= m - 1) {
+			struct sge_eth_rxq *rxq;
+			unsigned int id = __ffs(m) + i * BITS_PER_LONG;
+			struct sge_fl *fl = s->egr_map[id];
+
+			rxq = container_of(fl, struct sge_eth_rxq, fl);
+			if (spin_trylock_irq(&rxq->rspq.lock)) {
+				if (!napi_is_scheduled(&rxq->rspq.napi)) {
+					if (fl_starving(fl)) {
+						fl->starving++;
+						__refill_fl(adap, fl);
+					}
+					if (!fl_starving(fl))
+						clear_bit(id, s->starving_fl);
+				}
+				spin_unlock(&rxq->rspq.lock);
+			}
+		}
+
+	t4_write_reg(adap, SGE_DEBUG_INDEX, 13);
+	cnt[0] = t4_read_reg(adap, SGE_DEBUG_DATA_HIGH);
+	cnt[1] = t4_read_reg(adap, SGE_DEBUG_DATA_LOW);
+
+	for (i = 0; i < 2; i++)
+		if (cnt[i] >= s->starve_thres) {
+			if (s->idma_state[i])
+				continue;
+			s->idma_state[i] = 1;
+			t4_write_reg(adap, SGE_DEBUG_INDEX, 11);
+			m = t4_read_reg(adap, SGE_DEBUG_DATA_LOW) >> (i * 16);
+			dev_warn(adap->pdev_dev,
+				 "SGE idma%u starvation detected for "
+				 "queue %lu\n", i, m & 0xffff);
+		} else if (s->idma_state[i])
+			s->idma_state[i] = 0;
+
+	mod_timer(&s->rx_timer, jiffies + RX_QCHECK_PERIOD);
+}
+
+static void sge_tx_timer_cb(unsigned long data)
+{
+	unsigned long m, i;
+	struct adapter *adap = (struct adapter *)data;
+	struct sge *s = &adap->sge;
+
+	for (i = 0; i < ARRAY_SIZE(s->txq_maperr); i++)
+		for (m = s->txq_maperr[i]; m; m &= m - 1) {
+			unsigned long id = __ffs(m) + i * BITS_PER_LONG;
+			struct sge_ofld_txq *txq = s->egr_map[id];
+
+			clear_bit(id, s->txq_maperr);
+			tasklet_schedule(&txq->qresume_tsk);
+		}
+
+	mod_timer(&s->tx_timer, jiffies + TX_QCHECK_PERIOD);
+}
+
+int t4_sge_alloc_rxq(struct adapter *adap, struct sge_rspq *iq, bool fwevtq,
+		     struct net_device *dev, int intr_idx,
+		     struct sge_fl *fl, rspq_handler_t hnd)
+{
+	int ret, flsz = 0;
+	struct fw_iq_cmd c;
+	struct port_info *pi = netdev_priv(dev);
+
+	/* Size needs to be multiple of 16, including status entry. */
+	iq->size = roundup(iq->size, 16);
+
+	iq->desc = alloc_ring(adap->pdev_dev, iq->size, iq->iqe_len, 0,
+			      &iq->phys_addr, NULL, 0);
+	if (!iq->desc)
+		return -ENOMEM;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_IQ_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_WRITE | FW_CMD_EXEC |
+			    FW_IQ_CMD_PFN(0) | FW_IQ_CMD_VFN(0));
+	c.alloc_to_len16 = htonl(FW_IQ_CMD_ALLOC | FW_IQ_CMD_IQSTART(1) |
+				 FW_LEN16(c));
+	c.type_to_iqandstindex = htonl(FW_IQ_CMD_TYPE(FW_IQ_TYPE_FL_INT_CAP) |
+		FW_IQ_CMD_IQASYNCH(fwevtq) | FW_IQ_CMD_VIID(pi->viid) |
+		FW_IQ_CMD_IQANDST(intr_idx < 0) | FW_IQ_CMD_IQANUD(1) |
+		FW_IQ_CMD_IQANDSTINDEX(intr_idx >= 0 ? intr_idx :
+							-intr_idx - 1));
+	c.iqdroprss_to_iqesize = htons(FW_IQ_CMD_IQPCIECH(pi->tx_chan) |
+		FW_IQ_CMD_IQGTSMODE |
+		FW_IQ_CMD_IQINTCNTTHRESH(iq->pktcnt_idx) |
+		FW_IQ_CMD_IQESIZE(ilog2(iq->iqe_len) - 4));
+	c.iqsize = htons(iq->size);
+	c.iqaddr = cpu_to_be64(iq->phys_addr);
+
+	if (fl) {
+		fl->size = roundup(fl->size, 8);
+		fl->desc = alloc_ring(adap->pdev_dev, fl->size, sizeof(__be64),
+				      sizeof(struct rx_sw_desc), &fl->addr,
+				      &fl->sdesc, STAT_LEN);
+		if (!fl->desc)
+			goto fl_nomem;
+
+		flsz = fl->size / 8 + STAT_LEN / sizeof(struct tx_desc);
+		c.iqns_to_fl0congen = htonl(FW_IQ_CMD_FL0PACKEN |
+					    FW_IQ_CMD_FL0PADEN);
+		c.fl0dcaen_to_fl0cidxfthresh = htons(FW_IQ_CMD_FL0FBMIN(2) |
+				FW_IQ_CMD_FL0FBMAX(3));
+		c.fl0size = htons(flsz);
+		c.fl0addr = cpu_to_be64(fl->addr);
+	}
+
+	ret = t4_wr_mbox(adap, 0, &c, sizeof(c), &c);
+	if (ret)
+		goto err;
+
+	spin_lock_init(&iq->lock);
+	netif_napi_add(dev, &iq->napi, napi_rx_handler, 64);
+	iq->cur_desc = iq->desc;
+	iq->cidx = 0;
+	iq->gen = 1;
+	iq->next_intr_params = iq->intr_params;
+	iq->cntxt_id = ntohs(c.iqid);
+	iq->abs_id = ntohs(c.physiqid);
+	iq->size--;                           /* subtract status entry */
+	iq->adap = adap;
+	iq->netdev = dev;
+	iq->handler = hnd;
+
+	/* set offset to -1 to distinguish ingress queues without FL */
+	iq->offset = fl ? 0 : -1;
+
+	adap->sge.ingr_map[iq->cntxt_id] = iq;
+
+	if (fl) {
+		fl->cntxt_id = htons(c.fl0id);
+		fl->avail = fl->pend_cred = 0;
+		fl->pidx = fl->cidx = 0;
+		fl->alloc_failed = fl->large_alloc_failed = fl->starving = 0;
+		adap->sge.egr_map[fl->cntxt_id] = fl;
+		replenish_fl(adap, fl, GFP_KERNEL);
+	}
+	return 0;
+
+fl_nomem:
+	ret = -ENOMEM;
+err:
+	if (iq->desc) {
+		dma_free_coherent(adap->pdev_dev, iq->size * iq->iqe_len,
+				  iq->desc, iq->phys_addr);
+		iq->desc = NULL;
+	}
+	if (fl && fl->desc) {
+		kfree(fl->sdesc);
+		fl->sdesc = NULL;
+		dma_free_coherent(adap->pdev_dev, flsz * sizeof(struct tx_desc),
+				  fl->desc, fl->addr);
+		fl->desc = NULL;
+	}
+	return ret;
+}
+
+static void init_txq(struct adapter *adap, struct sge_txq *q, unsigned int id)
+{
+	q->in_use = 0;
+	q->cidx = q->pidx = 0;
+	q->stops = q->restarts = 0;
+	q->stat = (void *)&q->desc[q->size];
+	q->cntxt_id = id;
+	adap->sge.egr_map[id] = q;
+}
+
+int t4_sge_alloc_eth_txq(struct adapter *adap, struct sge_eth_txq *txq,
+			 struct net_device *dev, struct netdev_queue *netdevq,
+			 unsigned int iqid)
+{
+	int ret, nentries;
+	struct fw_eq_eth_cmd c;
+	struct port_info *pi = netdev_priv(dev);
+
+	/* Add status entries */
+	nentries = txq->q.size + STAT_LEN / sizeof(struct tx_desc);
+
+	txq->q.desc = alloc_ring(adap->pdev_dev, txq->q.size,
+			sizeof(struct tx_desc), sizeof(struct tx_sw_desc),
+			&txq->q.phys_addr, &txq->q.sdesc, STAT_LEN);
+	if (!txq->q.desc)
+		return -ENOMEM;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_EQ_ETH_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_WRITE | FW_CMD_EXEC |
+			    FW_EQ_ETH_CMD_PFN(0) | FW_EQ_ETH_CMD_VFN(0));
+	c.alloc_to_len16 = htonl(FW_EQ_ETH_CMD_ALLOC |
+				 FW_EQ_ETH_CMD_EQSTART | FW_LEN16(c));
+	c.viid_pkd = htonl(FW_EQ_ETH_CMD_VIID(pi->viid));
+	c.fetchszm_to_iqid = htonl(FW_EQ_ETH_CMD_HOSTFCMODE(2) |
+				   FW_EQ_ETH_CMD_PCIECHN(pi->tx_chan) |
+				   FW_EQ_ETH_CMD_IQID(iqid));
+	c.dcaen_to_eqsize = htonl(FW_EQ_ETH_CMD_FBMIN(2) |
+				  FW_EQ_ETH_CMD_FBMAX(3) |
+				  FW_EQ_ETH_CMD_CIDXFTHRESH(5) |
+				  FW_EQ_ETH_CMD_EQSIZE(nentries));
+	c.eqaddr = cpu_to_be64(txq->q.phys_addr);
+
+	ret = t4_wr_mbox(adap, 0, &c, sizeof(c), &c);
+	if (ret) {
+		kfree(txq->q.sdesc);
+		txq->q.sdesc = NULL;
+		dma_free_coherent(adap->pdev_dev,
+				  nentries * sizeof(struct tx_desc),
+				  txq->q.desc, txq->q.phys_addr);
+		txq->q.desc = NULL;
+		return ret;
+	}
+
+	init_txq(adap, &txq->q, FW_EQ_ETH_CMD_EQID_GET(ntohl(c.eqid_pkd)));
+	txq->txq = netdevq;
+	txq->tso = txq->tx_cso = txq->vlan_ins = 0;
+	txq->mapping_err = 0;
+	return 0;
+}
+
+int t4_sge_alloc_ctrl_txq(struct adapter *adap, struct sge_ctrl_txq *txq,
+			  struct net_device *dev, unsigned int iqid,
+			  unsigned int cmplqid)
+{
+	int ret, nentries;
+	struct fw_eq_ctrl_cmd c;
+	struct port_info *pi = netdev_priv(dev);
+
+	/* Add status entries */
+	nentries = txq->q.size + STAT_LEN / sizeof(struct tx_desc);
+
+	txq->q.desc = alloc_ring(adap->pdev_dev, nentries,
+				 sizeof(struct tx_desc), 0, &txq->q.phys_addr,
+				 NULL, 0);
+	if (!txq->q.desc)
+		return -ENOMEM;
+
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_EQ_CTRL_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_WRITE | FW_CMD_EXEC |
+			    FW_EQ_CTRL_CMD_PFN(0) | FW_EQ_CTRL_CMD_VFN(0));
+	c.alloc_to_len16 = htonl(FW_EQ_CTRL_CMD_ALLOC |
+				 FW_EQ_CTRL_CMD_EQSTART | FW_LEN16(c));
+	c.cmpliqid_eqid = htonl(FW_EQ_CTRL_CMD_CMPLIQID(cmplqid));
+	c.physeqid_pkd = htonl(0);
+	c.fetchszm_to_iqid = htonl(FW_EQ_CTRL_CMD_HOSTFCMODE(2) |
+				   FW_EQ_CTRL_CMD_PCIECHN(pi->tx_chan) |
+				   FW_EQ_CTRL_CMD_IQID(iqid));
+	c.dcaen_to_eqsize = htonl(FW_EQ_CTRL_CMD_FBMIN(2) |
+				  FW_EQ_CTRL_CMD_FBMAX(3) |
+				  FW_EQ_CTRL_CMD_CIDXFTHRESH(5) |
+				  FW_EQ_CTRL_CMD_EQSIZE(nentries));
+	c.eqaddr = cpu_to_be64(txq->q.phys_addr);
+
+	ret = t4_wr_mbox(adap, 0, &c, sizeof(c), &c);
+	if (ret) {
+		dma_free_coherent(adap->pdev_dev,
+				  nentries * sizeof(struct tx_desc),
+				  txq->q.desc, txq->q.phys_addr);
+		txq->q.desc = NULL;
+		return ret;
+	}
+
+	init_txq(adap, &txq->q, FW_EQ_CTRL_CMD_EQID_GET(ntohl(c.cmpliqid_eqid)));
+	txq->adap = adap;
+	skb_queue_head_init(&txq->sendq);
+	tasklet_init(&txq->qresume_tsk, restart_ctrlq, (unsigned long)txq);
+	txq->full = 0;
+	return 0;
+}
+
+int t4_sge_alloc_ofld_txq(struct adapter *adap, struct sge_ofld_txq *txq,
+			  struct net_device *dev, unsigned int iqid)
+{
+	int ret, nentries;
+	struct fw_eq_ofld_cmd c;
+	struct port_info *pi = netdev_priv(dev);
+
+	/* Add status entries */
+	nentries = txq->q.size + STAT_LEN / sizeof(struct tx_desc);
+
+	txq->q.desc = alloc_ring(adap->pdev_dev, txq->q.size,
+			sizeof(struct tx_desc), sizeof(struct tx_sw_desc),
+			&txq->q.phys_addr, &txq->q.sdesc, STAT_LEN);
+	if (!txq->q.desc)
+		return -ENOMEM;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_EQ_OFLD_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_WRITE | FW_CMD_EXEC |
+			    FW_EQ_OFLD_CMD_PFN(0) | FW_EQ_OFLD_CMD_VFN(0));
+	c.alloc_to_len16 = htonl(FW_EQ_OFLD_CMD_ALLOC |
+				 FW_EQ_OFLD_CMD_EQSTART | FW_LEN16(c));
+	c.fetchszm_to_iqid = htonl(FW_EQ_OFLD_CMD_HOSTFCMODE(2) |
+				   FW_EQ_OFLD_CMD_PCIECHN(pi->tx_chan) |
+				   FW_EQ_OFLD_CMD_IQID(iqid));
+	c.dcaen_to_eqsize = htonl(FW_EQ_OFLD_CMD_FBMIN(2) |
+				  FW_EQ_OFLD_CMD_FBMAX(3) |
+				  FW_EQ_OFLD_CMD_CIDXFTHRESH(5) |
+				  FW_EQ_OFLD_CMD_EQSIZE(nentries));
+	c.eqaddr = cpu_to_be64(txq->q.phys_addr);
+
+	ret = t4_wr_mbox(adap, 0, &c, sizeof(c), &c);
+	if (ret) {
+		kfree(txq->q.sdesc);
+		txq->q.sdesc = NULL;
+		dma_free_coherent(adap->pdev_dev,
+				  nentries * sizeof(struct tx_desc),
+				  txq->q.desc, txq->q.phys_addr);
+		txq->q.desc = NULL;
+		return ret;
+	}
+
+	init_txq(adap, &txq->q, FW_EQ_OFLD_CMD_EQID_GET(ntohl(c.eqid_pkd)));
+	txq->adap = adap;
+	skb_queue_head_init(&txq->sendq);
+	tasklet_init(&txq->qresume_tsk, restart_ofldq, (unsigned long)txq);
+	txq->full = 0;
+	txq->mapping_err = 0;
+	return 0;
+}
+
+static void free_txq(struct adapter *adap, struct sge_txq *q)
+{
+	dma_free_coherent(adap->pdev_dev,
+			  q->size * sizeof(struct tx_desc) + STAT_LEN,
+			  q->desc, q->phys_addr);
+	q->cntxt_id = 0;
+	q->sdesc = NULL;
+	q->desc = NULL;
+}
+
+static void free_rspq_fl(struct adapter *adap, struct sge_rspq *rq,
+			 struct sge_fl *fl)
+{
+	unsigned int fl_id = fl ? fl->cntxt_id : 0xffff;
+
+	t4_iq_free(adap, 0, 0, 0, FW_IQ_TYPE_FL_INT_CAP, rq->cntxt_id, fl_id,
+		   0xffff);
+	dma_free_coherent(adap->pdev_dev, (rq->size + 1) * rq->iqe_len,
+			  rq->desc, rq->phys_addr);
+	netif_napi_del(&rq->napi);
+	rq->netdev = NULL;
+	rq->cntxt_id = rq->abs_id = 0;
+	rq->desc = NULL;
+
+	if (fl) {
+		free_rx_bufs(adap, fl, fl->avail);
+		dma_free_coherent(adap->pdev_dev, fl->size * 8 + STAT_LEN,
+				  fl->desc, fl->addr);
+		kfree(fl->sdesc);
+		fl->sdesc = NULL;
+		fl->cntxt_id = 0;
+		fl->desc = NULL;
+	}
+}
+
+/**
+ *	t4_free_sge_resources - free SGE resources
+ *	@adap: the adapter
+ *
+ *	Frees resources used by the SGE queue sets.
+ */
+void t4_free_sge_resources(struct adapter *adap)
+{
+	int i;
+	struct sge_eth_rxq *eq = adap->sge.ethrxq;
+	struct sge_eth_txq *etq = adap->sge.ethtxq;
+	struct sge_ofld_rxq *oq = adap->sge.ofldrxq;
+	struct sge_rspq *intrq = &adap->sge.intrq;
+
+	/* clean up Ethernet Tx/Rx queues */
+	for (i = 0; i < adap->sge.ethqsets; i++, eq++, etq++) {
+		if (eq->rspq.desc)
+			free_rspq_fl(adap, &eq->rspq, &eq->fl);
+		if (etq->q.desc) {
+			t4_eth_eq_free(adap, 0, 0, 0, etq->q.cntxt_id);
+			free_tx_desc(adap, &etq->q, etq->q.in_use, true);
+			kfree(etq->q.sdesc);
+			free_txq(adap, &etq->q);
+		}
+	}
+
+	/* clean up RDMA and iSCSI Rx queues */
+	for (i = 0; i < adap->sge.ofldqsets; i++, oq++) {
+		if (oq->rspq.desc)
+			free_rspq_fl(adap, &oq->rspq, &oq->fl);
+	}
+	for (i = 0, oq = adap->sge.rdmarxq; i < adap->sge.rdmaqs; i++, oq++) {
+		if (oq->rspq.desc)
+			free_rspq_fl(adap, &oq->rspq, &oq->fl);
+	}
+
+	/* clean up offload Tx queues */
+	for (i = 0; i < ARRAY_SIZE(adap->sge.ofldtxq); i++) {
+		struct sge_ofld_txq *q = &adap->sge.ofldtxq[i];
+
+		if (q->q.desc) {
+			tasklet_kill(&q->qresume_tsk);
+			t4_ofld_eq_free(adap, 0, 0, 0, q->q.cntxt_id);
+			free_tx_desc(adap, &q->q, q->q.in_use, false);
+			kfree(q->q.sdesc);
+			__skb_queue_purge(&q->sendq);
+			free_txq(adap, &q->q);
+		}
+	}
+
+	/* clean up control Tx queues */
+	for (i = 0; i < ARRAY_SIZE(adap->sge.ctrlq); i++) {
+		struct sge_ctrl_txq *cq = &adap->sge.ctrlq[i];
+
+		if (cq->q.desc) {
+			tasklet_kill(&cq->qresume_tsk);
+			t4_ctrl_eq_free(adap, 0, 0, 0, cq->q.cntxt_id);
+			__skb_queue_purge(&cq->sendq);
+			free_txq(adap, &cq->q);
+		}
+	}
+
+	if (adap->sge.fw_evtq.desc)
+		free_rspq_fl(adap, &adap->sge.fw_evtq, NULL);
+
+	if (intrq->desc) {
+		t4_iq_free(adap, 0, 0, 0, FW_IQ_TYPE_FL_INT_CAP,
+			   intrq->cntxt_id, 0xffff, 0xffff);
+		dma_free_coherent(adap->pdev_dev,
+				  (intrq->size + 1) * intrq->iqe_len,
+				  intrq->desc, intrq->phys_addr);
+		intrq->cntxt_id = intrq->abs_id = 0;
+		intrq->desc = NULL;
+		/* this queue doesn't use NAPI */
+	}
+}
+
+void t4_sge_start(struct adapter *adap)
+{
+	mod_timer(&adap->sge.rx_timer, jiffies + RX_QCHECK_PERIOD);
+	mod_timer(&adap->sge.tx_timer, jiffies + TX_QCHECK_PERIOD);
+}
+
+/**
+ *	t4_sge_stop - disable SGE operation
+ *	@adap: the adapter
+ *
+ *	Stop tasklets and timers associated with the DMA engine.  Note that
+ *	this is effective only if measures have been taken to disable any HW
+ *	events that may restart them.
+ */
+void t4_sge_stop(struct adapter *adap)
+{
+	int i;
+	struct sge *s = &adap->sge;
+
+	if (in_interrupt())  /* actions below require waiting */
+		return;
+
+	if (s->rx_timer.function)
+		del_timer_sync(&s->rx_timer);
+	if (s->tx_timer.function)
+		del_timer_sync(&s->tx_timer);
+
+	for (i = 0; i < ARRAY_SIZE(s->ofldtxq); i++) {
+		struct sge_ofld_txq *q = &s->ofldtxq[i];
+
+		if (q->q.desc)
+			tasklet_kill(&q->qresume_tsk);
+	}
+	for (i = 0; i < ARRAY_SIZE(s->ctrlq); i++) {
+		struct sge_ctrl_txq *cq = &s->ctrlq[i];
+
+		if (cq->q.desc)
+			tasklet_kill(&cq->qresume_tsk);
+	}
+}
+
+/**
+ *	t4_sge_init - initialize SGE
+ *	@adap: the adapter
+ *
+ *	Performs SGE initialization needed every time after a chip reset.
+ *	We do not initialize any of the queues here, instead the driver
+ *	top-level must request them individually.
+ */
+void t4_sge_init(struct adapter *adap)
+{
+	struct sge *s = &adap->sge;
+	unsigned int fl_align_log = ilog2(FL_ALIGN);
+
+	t4_set_reg_field(adap, SGE_CONTROL, PKTSHIFT_MASK |
+			 INGPADBOUNDARY_MASK | EGRSTATUSPAGESIZE,
+			 INGPADBOUNDARY(fl_align_log - 5) | PKTSHIFT(2) |
+			 RXPKTCPLMODE |
+			 (STAT_LEN == 128 ? EGRSTATUSPAGESIZE : 0));
+	t4_set_reg_field(adap, SGE_HOST_PAGE_SIZE, HOSTPAGESIZEPF0_MASK,
+			 HOSTPAGESIZEPF0(PAGE_SHIFT - 10));
+	t4_write_reg(adap, SGE_FL_BUFFER_SIZE0, PAGE_SIZE);
+#if FL_PG_ORDER > 0
+	t4_write_reg(adap, SGE_FL_BUFFER_SIZE1, PAGE_SIZE << FL_PG_ORDER);
+#endif
+	t4_write_reg(adap, SGE_INGRESS_RX_THRESHOLD,
+		     THRESHOLD_0(s->counter_val[0]) |
+		     THRESHOLD_1(s->counter_val[1]) |
+		     THRESHOLD_2(s->counter_val[2]) |
+		     THRESHOLD_3(s->counter_val[3]));
+	t4_write_reg(adap, SGE_TIMER_VALUE_0_AND_1,
+		     TIMERVALUE0(us_to_core_ticks(adap, s->timer_val[0])) |
+		     TIMERVALUE1(us_to_core_ticks(adap, s->timer_val[1])));
+	t4_write_reg(adap, SGE_TIMER_VALUE_2_AND_3,
+		     TIMERVALUE0(us_to_core_ticks(adap, s->timer_val[2])) |
+		     TIMERVALUE1(us_to_core_ticks(adap, s->timer_val[3])));
+	t4_write_reg(adap, SGE_TIMER_VALUE_4_AND_5,
+		     TIMERVALUE0(us_to_core_ticks(adap, s->timer_val[4])) |
+		     TIMERVALUE1(us_to_core_ticks(adap, s->timer_val[5])));
+	setup_timer(&s->rx_timer, sge_rx_timer_cb, (unsigned long)adap);
+	setup_timer(&s->tx_timer, sge_tx_timer_cb, (unsigned long)adap);
+	s->starve_thres = core_ticks_per_usec(adap) * 1000000;  /* 1 s */
+	s->idma_state[0] = s->idma_state[1] = 0;
+}
-- 
1.5.4

Re: [PATCH 4/7] cxgb4: Add packet queues and packet DMA code

[Stephen Hemminger]

On Fri, 26 Feb 2010 15:35:36 -0800
Dimitris Michailidis <dm@chelsio.com> wrote:

> +
> +/**
> + *	need_skb_unmap - does the platform need unmapping of sk_buffs?
> + *
> + *	Returns true if the platfrom needs sk_buff unmapping.  The compiler
> + *	optimizes away unecessary code if this returns true.
> + */
> +static inline int need_skb_unmap(void)
> +{
> +	/*
> +	 * This structure is used to tell if the platfrom needs buffer
> +	 * unmapping by checking if DECLARE_PCI_UNMAP_ADDR defines anything.
> +	 */
> +	struct dummy {
> +		DECLARE_PCI_UNMAP_ADDR(addr);
> +	};
> +
> +	return sizeof(struct dummy) != 0;
> +}
> +

I would prefer one code path and let the compiler decide if unmap
should be nop; rather than this kind of trick code.

-- 

RE: [PATCH 4/7] cxgb4: Add packet queues and packet DMA code

[Dimitrios Michailidis]

> -----Original Message-----
> From: Stephen Hemminger [mailto:shemminger@vyatta.com]
> Sent: Friday, February 26, 2010 3:58 PM
> To: Dimitrios Michailidis
> Cc: netdev@vger.kernel.org
> Subject: Re: [PATCH 4/7] cxgb4: Add packet queues and packet DMA code
> 
> On Fri, 26 Feb 2010 15:35:36 -0800
> Dimitris Michailidis <dm@chelsio.com> wrote:
> 
> > +
> > +/**
> > + *	need_skb_unmap - does the platform need unmapping of sk_buffs?
> > + *
> > + *	Returns true if the platfrom needs sk_buff unmapping.  The
compiler
> > + *	optimizes away unecessary code if this returns true.
> > + */
> > +static inline int need_skb_unmap(void)
> > +{
> > +	/*
> > +	 * This structure is used to tell if the platfrom needs buffer
> > +	 * unmapping by checking if DECLARE_PCI_UNMAP_ADDR defines
anything.
> > +	 */
> > +	struct dummy {
> > +		DECLARE_PCI_UNMAP_ADDR(addr);
> > +	};
> > +
> > +	return sizeof(struct dummy) != 0;
> > +}
> > +
> 
> I would prefer one code path and let the compiler decide if unmap
> should be nop; rather than this kind of trick code.

I believe this function does not generate any code, the compiler
statically figures out the result and optimizes any conditionals that
call it.  What option do you have in mind that would tell the compiler
if unmap is nop?

Re: [PATCH 4/7] cxgb4: Add packet queues and packet DMA code

[Stephen Hemminger]

On Fri, 26 Feb 2010 16:10:07 -0800
"Dimitrios Michailidis" <dm@chelsio.com> wrote:

> 
> 
> > -----Original Message-----
> > From: Stephen Hemminger [mailto:shemminger@vyatta.com]
> > Sent: Friday, February 26, 2010 3:58 PM
> > To: Dimitrios Michailidis
> > Cc: netdev@vger.kernel.org
> > Subject: Re: [PATCH 4/7] cxgb4: Add packet queues and packet DMA code
> > 
> > On Fri, 26 Feb 2010 15:35:36 -0800
> > Dimitris Michailidis <dm@chelsio.com> wrote:
> > 
> > > +
> > > +/**
> > > + *	need_skb_unmap - does the platform need unmapping of sk_buffs?
> > > + *
> > > + *	Returns true if the platfrom needs sk_buff unmapping.  The
> compiler
> > > + *	optimizes away unecessary code if this returns true.
> > > + */
> > > +static inline int need_skb_unmap(void)
> > > +{
> > > +	/*
> > > +	 * This structure is used to tell if the platfrom needs buffer
> > > +	 * unmapping by checking if DECLARE_PCI_UNMAP_ADDR defines
> anything.
> > > +	 */
> > > +	struct dummy {
> > > +		DECLARE_PCI_UNMAP_ADDR(addr);
> > > +	};
> > > +
> > > +	return sizeof(struct dummy) != 0;
> > > +}
> > > +
> > 
> > I would prefer one code path and let the compiler decide if unmap
> > should be nop; rather than this kind of trick code.
> 
> I believe this function does not generate any code, the compiler
> statically figures out the result and optimizes any conditionals that
> call it.  What option do you have in mind that would tell the compiler
> if unmap is nop?

Why do you care, let compiler decide.

-- 

RE: [PATCH 4/7] cxgb4: Add packet queues and packet DMA code

[Dimitrios Michailidis]

> -----Original Message-----
> From: netdev-owner@vger.kernel.org
[mailto:netdev-owner@vger.kernel.org]
> On Behalf Of Stephen Hemminger
> Sent: Friday, February 26, 2010 4:16 PM
> To: Dimitrios Michailidis
> Cc: netdev@vger.kernel.org
> Subject: Re: [PATCH 4/7] cxgb4: Add packet queues and packet DMA code
> 
> On Fri, 26 Feb 2010 16:10:07 -0800
> "Dimitrios Michailidis" <dm@chelsio.com> wrote:
> 
> >
> >
> > > -----Original Message-----
> > > From: Stephen Hemminger [mailto:shemminger@vyatta.com]
> > > Sent: Friday, February 26, 2010 3:58 PM
> > > To: Dimitrios Michailidis
> > > Cc: netdev@vger.kernel.org
> > > Subject: Re: [PATCH 4/7] cxgb4: Add packet queues and packet DMA
code
> > >
> > > On Fri, 26 Feb 2010 15:35:36 -0800
> > > Dimitris Michailidis <dm@chelsio.com> wrote:
> > >
> > > > +
> > > > +/**
> > > > + *	need_skb_unmap - does the platform need unmapping of
sk_buffs?
> > > > + *
> > > > + *	Returns true if the platfrom needs sk_buff unmapping.
The
> > compiler
> > > > + *	optimizes away unecessary code if this returns true.
> > > > + */
> > > > +static inline int need_skb_unmap(void)
> > > > +{
> > > > +	/*
> > > > +	 * This structure is used to tell if the platfrom needs
buffer
> > > > +	 * unmapping by checking if DECLARE_PCI_UNMAP_ADDR
defines
> > anything.
> > > > +	 */
> > > > +	struct dummy {
> > > > +		DECLARE_PCI_UNMAP_ADDR(addr);
> > > > +	};
> > > > +
> > > > +	return sizeof(struct dummy) != 0;
> > > > +}
> > > > +
> > >
> > > I would prefer one code path and let the compiler decide if unmap
> > > should be nop; rather than this kind of trick code.
> >
> > I believe this function does not generate any code, the compiler
> > statically figures out the result and optimizes any conditionals
that
> > call it.  What option do you have in mind that would tell the
compiler
> > if unmap is nop?
> 
> Why do you care, let compiler decide.

I want the compiler to decide but I must be missing something.  How is
the compiler going to decide?  The purpose of this code is to tell the
compiler if unmapping is nop.  Does it have another way to tell?

Re: [PATCH 4/7] cxgb4: Add packet queues and packet DMA code

[Stephen Hemminger]

On Fri, 26 Feb 2010 16:52:10 -0800
"Dimitrios Michailidis" <dm@chelsio.com> wrote:

> 
> 
> > -----Original Message-----
> > From: netdev-owner@vger.kernel.org
> [mailto:netdev-owner@vger.kernel.org]
> > On Behalf Of Stephen Hemminger
> > Sent: Friday, February 26, 2010 4:16 PM
> > To: Dimitrios Michailidis
> > Cc: netdev@vger.kernel.org
> > Subject: Re: [PATCH 4/7] cxgb4: Add packet queues and packet DMA code
> > 
> > On Fri, 26 Feb 2010 16:10:07 -0800
> > "Dimitrios Michailidis" <dm@chelsio.com> wrote:
> > 
> > >
> > >
> > > > -----Original Message-----
> > > > From: Stephen Hemminger [mailto:shemminger@vyatta.com]
> > > > Sent: Friday, February 26, 2010 3:58 PM
> > > > To: Dimitrios Michailidis
> > > > Cc: netdev@vger.kernel.org
> > > > Subject: Re: [PATCH 4/7] cxgb4: Add packet queues and packet DMA
> code
> > > >
> > > > On Fri, 26 Feb 2010 15:35:36 -0800
> > > > Dimitris Michailidis <dm@chelsio.com> wrote:
> > > >
> > > > > +
> > > > > +/**
> > > > > + *	need_skb_unmap - does the platform need unmapping of
> sk_buffs?
> > > > > + *
> > > > > + *	Returns true if the platfrom needs sk_buff unmapping.
> The
> > > compiler
> > > > > + *	optimizes away unecessary code if this returns true.
> > > > > + */
> > > > > +static inline int need_skb_unmap(void)
> > > > > +{
> > > > > +	/*
> > > > > +	 * This structure is used to tell if the platfrom needs
> buffer
> > > > > +	 * unmapping by checking if DECLARE_PCI_UNMAP_ADDR
> defines
> > > anything.
> > > > > +	 */
> > > > > +	struct dummy {
> > > > > +		DECLARE_PCI_UNMAP_ADDR(addr);
> > > > > +	};
> > > > > +
> > > > > +	return sizeof(struct dummy) != 0;
> > > > > +}
> > > > > +
> > > >
> > > > I would prefer one code path and let the compiler decide if unmap
> > > > should be nop; rather than this kind of trick code.
> > >
> > > I believe this function does not generate any code, the compiler
> > > statically figures out the result and optimizes any conditionals
> that
> > > call it.  What option do you have in mind that would tell the
> compiler
> > > if unmap is nop?
> > 
> > Why do you care, let compiler decide.
> 
> I want the compiler to decide but I must be missing something.  How is
> the compiler going to decide?  The purpose of this code is to tell the
> compiler if unmapping is nop.  Does it have another way to tell?

I looked closer, it is all tied up with deferred skb unmapping and that
code looks inherited from cxgb3. Not sure why you need to hold to skb for
so long: some offload requirement? Anyway keep it they way it is..

-- 

RE: [PATCH 4/7] cxgb4: Add packet queues and packet DMA code

[Dimitrios Michailidis]

> -----Original Message-----
> From: Stephen Hemminger [mailto:shemminger@vyatta.com]
> Sent: Friday, February 26, 2010 5:31 PM
> To: Dimitrios Michailidis
> Cc: netdev@vger.kernel.org
> Subject: Re: [PATCH 4/7] cxgb4: Add packet queues and packet DMA code
> 
> On Fri, 26 Feb 2010 16:52:10 -0800
> "Dimitrios Michailidis" <dm@chelsio.com> wrote:
> 
> >
> >
> > > -----Original Message-----
> > > From: netdev-owner@vger.kernel.org
> > [mailto:netdev-owner@vger.kernel.org]
> > > On Behalf Of Stephen Hemminger
> > > Sent: Friday, February 26, 2010 4:16 PM
> > > To: Dimitrios Michailidis
> > > Cc: netdev@vger.kernel.org
> > > Subject: Re: [PATCH 4/7] cxgb4: Add packet queues and packet DMA
code
> > >
> > > On Fri, 26 Feb 2010 16:10:07 -0800
> > > "Dimitrios Michailidis" <dm@chelsio.com> wrote:
> > >
> > > >
> > > >
> > > > > -----Original Message-----
> > > > > From: Stephen Hemminger [mailto:shemminger@vyatta.com]
> > > > > Sent: Friday, February 26, 2010 3:58 PM
> > > > > To: Dimitrios Michailidis
> > > > > Cc: netdev@vger.kernel.org
> > > > > Subject: Re: [PATCH 4/7] cxgb4: Add packet queues and packet
DMA
> > code
> > > > >
> > > > > On Fri, 26 Feb 2010 15:35:36 -0800
> > > > > Dimitris Michailidis <dm@chelsio.com> wrote:
> > > > >
> > > > > > +
> > > > > > +/**
> > > > > > + *	need_skb_unmap - does the platform need unmapping of
> > sk_buffs?
> > > > > > + *
> > > > > > + *	Returns true if the platfrom needs sk_buff unmapping.
> > The
> > > > compiler
> > > > > > + *	optimizes away unecessary code if this returns true.
> > > > > > + */
> > > > > > +static inline int need_skb_unmap(void)
> > > > > > +{
> > > > > > +	/*
> > > > > > +	 * This structure is used to tell if the platfrom needs
> > buffer
> > > > > > +	 * unmapping by checking if DECLARE_PCI_UNMAP_ADDR
> > defines
> > > > anything.
> > > > > > +	 */
> > > > > > +	struct dummy {
> > > > > > +		DECLARE_PCI_UNMAP_ADDR(addr);
> > > > > > +	};
> > > > > > +
> > > > > > +	return sizeof(struct dummy) != 0;
> > > > > > +}
> > > > > > +
> > > > >
> > > > > I would prefer one code path and let the compiler decide if
unmap
> > > > > should be nop; rather than this kind of trick code.
> > > >
> > > > I believe this function does not generate any code, the compiler
> > > > statically figures out the result and optimizes any conditionals
> > that
> > > > call it.  What option do you have in mind that would tell the
> > compiler
> > > > if unmap is nop?
> > >
> > > Why do you care, let compiler decide.
> >
> > I want the compiler to decide but I must be missing something.  How
is
> > the compiler going to decide?  The purpose of this code is to tell
the
> > compiler if unmapping is nop.  Does it have another way to tell?
> 
> I looked closer, it is all tied up with deferred skb unmapping and
that
> code looks inherited from cxgb3. Not sure why you need to hold to skb
for
> so long: some offload requirement? Anyway keep it they way it is..

Yes, offload packets can have their payload DMAd after their DMA
descriptor is freed and so they need to remain mapped longer.

Re: [PATCH 4/7] cxgb4: Add packet queues and packet DMA code

[David Miller]

From: Stephen Hemminger <shemminger@vyatta.com>
Date: Fri, 26 Feb 2010 15:57:56 -0800

> On Fri, 26 Feb 2010 15:35:36 -0800
> Dimitris Michailidis <dm@chelsio.com> wrote:
> 
>> +
>> +/**
>> + *	need_skb_unmap - does the platform need unmapping of sk_buffs?
>> + *
>> + *	Returns true if the platfrom needs sk_buff unmapping.  The compiler
>> + *	optimizes away unecessary code if this returns true.
>> + */
>> +static inline int need_skb_unmap(void)
>> +{
>> +	/*
>> +	 * This structure is used to tell if the platfrom needs buffer
>> +	 * unmapping by checking if DECLARE_PCI_UNMAP_ADDR defines anything.
>> +	 */
>> +	struct dummy {
>> +		DECLARE_PCI_UNMAP_ADDR(addr);
>> +	};
>> +
>> +	return sizeof(struct dummy) != 0;
>> +}
>> +
> 
> I would prefer one code path and let the compiler decide if unmap
> should be nop; rather than this kind of trick code.

Agreed, this is rediculious.  And there is currently no platform that
matters where this will trigger.

With IOMMUs and even the swiommu layer, even bog standard x86 systems
need the unmap addresses.

Therefore optimizing this is totally unreasonable.  Just use the provided
APIs and let the compiler and the platform defines do the work.

I'm amazed at how people do stuff like this.  If it's a reasonable
opimization, maybe, JUST MAYBE, it's better to optimize it generically
so that every driver gets the benefit NOT JUST YOUR individual driver.

Re: [PATCH 4/7] cxgb4: Add packet queues and packet DMA code

[David Miller]

From: "Dimitrios Michailidis" <dm@chelsio.com>
Date: Fri, 26 Feb 2010 16:10:07 -0800

> I believe this function does not generate any code, the compiler
> statically figures out the result and optimizes any conditionals that
> call it.  What option do you have in mind that would tell the compiler
> if unmap is nop?

I've got better questions:

1) Why is your driver so damn special?

   If this optimization is useful, it dosn't belong privately in
   some driver, it belongs in some generic spot.

2) What configuration does this even help for?  Even bog standard x86
   and x86_64 uses IOMMUs and thus make use of the unmap address.

   I cannot think of one platform that matters where this will even
   trigger.

Get rid of this junk, please.  Because even if it's valid, it doesn't
belong privately in your driver.

Re: [PATCH 4/7] cxgb4: Add packet queues and packet DMA code

[Dimitris Michailidis]

David Miller wrote:
> From: "Dimitrios Michailidis" <dm@chelsio.com>
> Date: Fri, 26 Feb 2010 16:10:07 -0800
> 
>> I believe this function does not generate any code, the compiler
>> statically figures out the result and optimizes any conditionals that
>> call it.  What option do you have in mind that would tell the compiler
>> if unmap is nop?
> 
> I've got better questions:
> 
> 1) Why is your driver so damn special?
> 
>    If this optimization is useful, it dosen't belong privately in
>    some driver, it belongs in some generic spot.

I agree that it would be better if this were available at some generic 
place, eg an arch header.  It is implicitly available for most arches 
through the definition of DECLARE_PCI_UNMAP_ADDR and related macros, which 
is why the driver is looking at that, but it would be nicer if it were 
explicitly available.  Among the arches MIPS has an explicit indication 
through its CONFIG_DMA_NEED_PCI_MAP_STATE but it's the only one.

> 
> 2) What configuration does this even help for?  Even bog standard x86
>    and x86_64 uses IOMMUs and thus make use of the unmap address.
> 
>    I cannot think of one platform that matters where this will even
>    trigger.

While x86_32 has configurations that don't use unmapping (see 
arch/x86/include/asm/pci.h) it is more helpful with less powerful CPUs, like 
PPC or MIPS one finds in embedded systems.  People use this device with 
embedded CPUs.

> 
> Get rid of this junk, please.  Because even if it's valid, it doesn't
> belong privately in your driver.
> 

I'd love not to have this in the driver but I don't know how to get the same 
result without it with the current state of affairs.

[PATCH 4/7] cxgb4: Add packet queues and packet DMA code

[Dimitris Michailidis]

Signed-off-by: Dimitris Michailidis <dm@chelsio.com>
---
 drivers/net/cxgb4/sge.c | 2438 +++++++++++++++++++++++++++++++++++++++++++++++
 1 files changed, 2438 insertions(+), 0 deletions(-)
 create mode 100644 drivers/net/cxgb4/sge.c

diff --git a/drivers/net/cxgb4/sge.c b/drivers/net/cxgb4/sge.c
new file mode 100644
index 0000000..eb81d73
--- /dev/null
+++ b/drivers/net/cxgb4/sge.c
@@ -0,0 +1,2438 @@
+/*
+ * This file is part of the Chelsio T4 Ethernet driver for Linux.
+ *
+ * Copyright (c) 2003-2010 Chelsio Communications, Inc. All rights reserved.
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and/or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+
+#include <linux/skbuff.h>
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/if_vlan.h>
+#include <linux/ip.h>
+#include <linux/dma-mapping.h>
+#include <net/ipv6.h>
+#include <net/tcp.h>
+#include "cxgb4.h"
+#include "t4_regs.h"
+#include "t4_msg.h"
+#include "t4fw_api.h"
+
+/*
+ * Rx buffer size.  We use largish buffers if possible but settle for single
+ * pages under memory shortage.
+ */
+#if PAGE_SHIFT >= 16
+# define FL_PG_ORDER 0
+#else
+# define FL_PG_ORDER (16 - PAGE_SHIFT)
+#endif
+
+/* RX_PULL_LEN should be <= RX_COPY_THRES */
+#define RX_COPY_THRES    256
+#define RX_PULL_LEN      128
+
+/*
+ * Main body length for sk_buffs used for Rx Ethernet packets with fragments.
+ * Should be >= RX_PULL_LEN but possibly bigger to give pskb_may_pull some room.
+ */
+#define RX_PKT_SKB_LEN   512
+
+/* Ethernet header padding prepended to RX_PKTs */
+#define RX_PKT_PAD 2
+
+/*
+ * Max number of Tx descriptors we clean up at a time.  Should be modest as
+ * freeing skbs isn't cheap and it happens while holding locks.  We just need
+ * to free packets faster than they arrive, we eventually catch up and keep
+ * the amortized cost reasonable.  Must be >= 2 * TXQ_STOP_THRES.
+ */
+#define MAX_TX_RECLAIM 16
+
+/*
+ * Max number of Rx buffers we replenish at a time.  Again keep this modest,
+ * allocating buffers isn't cheap either.
+ */
+#define MAX_RX_REFILL 16U
+
+/*
+ * Period of the Rx queue check timer.  This timer is infrequent as it has
+ * something to do only when the system experiences severe memory shortage.
+ */
+#define RX_QCHECK_PERIOD (HZ / 2)
+
+/*
+ * Period of the Tx queue check timer.
+ */
+#define TX_QCHECK_PERIOD (HZ / 2)
+
+/*
+ * Timer index used when backing off due to memory shortage.
+ */
+#define NOMEM_TMR_IDX (SGE_NTIMERS - 1)
+
+/*
+ * An FL with <= FL_STARVE_THRES buffers is starving and a periodic timer will
+ * attempt to refill it.
+ */
+#define FL_STARVE_THRES 4
+
+/*
+ * Suspend an Ethernet Tx queue with fewer available descriptors than this.
+ * This is the same as calc_tx_descs() for a TSO packet with
+ * nr_frags == MAX_SKB_FRAGS.
+ */
+#define ETHTXQ_STOP_THRES \
+	(1 + DIV_ROUND_UP((3 * MAX_SKB_FRAGS) / 2 + (MAX_SKB_FRAGS & 1), 8))
+
+/*
+ * Suspension threshold for non-Ethernet Tx queues.  We require enough room
+ * for a full sized WR.
+ */
+#define TXQ_STOP_THRES (SGE_MAX_WR_LEN / sizeof(struct tx_desc))
+
+/*
+ * Max Tx descriptor space we allow for an Ethernet packet to be inlined
+ * into a WR.
+ */
+#define MAX_IMM_TX_PKT_LEN 128
+
+/*
+ * Max size of a WR sent through a control Tx queue.
+ */
+#define MAX_CTRL_WR_LEN SGE_MAX_WR_LEN
+
+enum {
+	/* packet alignment in FL buffers */
+	FL_ALIGN = L1_CACHE_BYTES < 32 ? 32 : L1_CACHE_BYTES,
+	/* egress status entry size */
+	STAT_LEN = L1_CACHE_BYTES > 64 ? 128 : 64
+};
+
+struct tx_sw_desc {                /* SW state per Tx descriptor */
+	struct sk_buff *skb;
+	struct ulptx_sgl *sgl;
+};
+
+struct rx_sw_desc {                /* SW state per Rx descriptor */
+	struct page *page;
+	dma_addr_t dma_addr;
+};
+
+/*
+ * The low bits of rx_sw_desc.dma_addr have special meaning.
+ */
+enum {
+	RX_LARGE_BUF    = 1 << 0, /* buffer is larger than PAGE_SIZE */
+	RX_UNMAPPED_BUF = 1 << 1, /* buffer is not mapped */
+};
+
+static inline dma_addr_t get_buf_addr(const struct rx_sw_desc *d)
+{
+	return d->dma_addr & ~(dma_addr_t)(RX_LARGE_BUF | RX_UNMAPPED_BUF);
+}
+
+static inline bool is_buf_mapped(const struct rx_sw_desc *d)
+{
+	return !(d->dma_addr & RX_UNMAPPED_BUF);
+}
+
+/**
+ *	txq_avail - return the number of available slots in a Tx queue
+ *	@q: the Tx queue
+ *
+ *	Returns the number of descriptors in a Tx queue available to write new
+ *	packets.
+ */
+static inline unsigned int txq_avail(const struct sge_txq *q)
+{
+	return q->size - 1 - q->in_use;
+}
+
+/**
+ *	fl_cap - return the capacity of a free-buffer list
+ *	@fl: the FL
+ *
+ *	Returns the capacity of a free-buffer list.  The capacity is less than
+ *	the size because one descriptor needs to be left unpopulated, otherwise
+ *	HW will think the FL is empty.
+ */
+static inline unsigned int fl_cap(const struct sge_fl *fl)
+{
+	return fl->size - 8;   /* 1 descriptor = 8 buffers */
+}
+
+static inline bool fl_starving(const struct sge_fl *fl)
+{
+	return fl->avail - fl->pend_cred <= FL_STARVE_THRES;
+}
+
+static int map_skb(struct device *dev, const struct sk_buff *skb,
+		   dma_addr_t *addr)
+{
+	const skb_frag_t *fp, *end;
+	const struct skb_shared_info *si;
+
+	*addr = dma_map_single(dev, skb->data, skb_headlen(skb), DMA_TO_DEVICE);
+	if (dma_mapping_error(dev, *addr))
+		goto out_err;
+
+	si = skb_shinfo(skb);
+	end = &si->frags[si->nr_frags];
+
+	for (fp = si->frags; fp < end; fp++) {
+		*++addr = dma_map_page(dev, fp->page, fp->page_offset, fp->size,
+				       DMA_TO_DEVICE);
+		if (dma_mapping_error(dev, *addr))
+			goto unwind;
+	}
+	return 0;
+
+unwind:
+	while (fp-- > si->frags)
+		dma_unmap_page(dev, *--addr, fp->size, DMA_TO_DEVICE);
+
+	dma_unmap_single(dev, addr[-1], skb_headlen(skb), DMA_TO_DEVICE);
+out_err:
+	return -ENOMEM;
+}
+
+static void unmap_skb(struct device *dev, const struct sk_buff *skb,
+		      const dma_addr_t *addr)
+{
+	const skb_frag_t *fp, *end;
+	const struct skb_shared_info *si;
+
+	dma_unmap_single(dev, *addr++, skb_headlen(skb), DMA_TO_DEVICE);
+
+	si = skb_shinfo(skb);
+	end = &si->frags[si->nr_frags];
+	for (fp = si->frags; fp < end; fp++)
+		dma_unmap_page(dev, *addr++, fp->size, DMA_TO_DEVICE);
+}
+
+static void unmap_sgl(struct device *dev, const struct sk_buff *skb,
+		      const struct ulptx_sgl *sgl, const struct sge_txq *q)
+{
+	const struct ulptx_sge_pair *p;
+	unsigned int nfrags = skb_shinfo(skb)->nr_frags;
+
+	if (likely(skb_headlen(skb)))
+		dma_unmap_single(dev, be64_to_cpu(sgl->addr0), ntohl(sgl->len0),
+				 DMA_TO_DEVICE);
+	else {
+		dma_unmap_page(dev, be64_to_cpu(sgl->addr0), ntohl(sgl->len0),
+			       DMA_TO_DEVICE);
+		nfrags--;
+	}
+
+	/*
+	 * the complexity below is because of the possibility of a wrap-around
+	 * in the middle of an SGL
+	 */
+	for (p = sgl->sge; nfrags >= 2; nfrags -= 2) {
+		if (likely((u8 *)(p + 1) <= (u8 *)q->stat)) {
+unmap:			dma_unmap_page(dev, be64_to_cpu(p->addr[0]),
+				       ntohl(p->len[0]), DMA_TO_DEVICE);
+			dma_unmap_page(dev, be64_to_cpu(p->addr[1]),
+				       ntohl(p->len[1]), DMA_TO_DEVICE);
+			p++;
+		} else if ((u8 *)p == (u8 *)q->stat) {
+			p = (const struct ulptx_sge_pair *)q->desc;
+			goto unmap;
+		} else if ((u8 *)p + 8 == (u8 *)q->stat) {
+			const __be64 *addr = (const __be64 *)q->desc;
+
+			dma_unmap_page(dev, be64_to_cpu(addr[0]),
+				       ntohl(p->len[0]), DMA_TO_DEVICE);
+			dma_unmap_page(dev, be64_to_cpu(addr[1]),
+				       ntohl(p->len[1]), DMA_TO_DEVICE);
+			p = (const struct ulptx_sge_pair *)&addr[2];
+		} else {
+			const __be64 *addr = (const __be64 *)q->desc;
+
+			dma_unmap_page(dev, be64_to_cpu(p->addr[0]),
+				       ntohl(p->len[0]), DMA_TO_DEVICE);
+			dma_unmap_page(dev, be64_to_cpu(addr[0]),
+				       ntohl(p->len[1]), DMA_TO_DEVICE);
+			p = (const struct ulptx_sge_pair *)&addr[1];
+		}
+	}
+	if (nfrags) {
+		__be64 addr;
+
+		if ((u8 *)p == (u8 *)q->stat)
+			p = (const struct ulptx_sge_pair *)q->desc;
+		addr = (u8 *)p + 16 <= (u8 *)q->stat ? p->addr[0] :
+						       *(const __be64 *)q->desc;
+		dma_unmap_page(dev, be64_to_cpu(addr), ntohl(p->len[0]),
+			       DMA_TO_DEVICE);
+	}
+}
+
+/**
+ *	free_tx_desc - reclaims Tx descriptors and their buffers
+ *	@adapter: the adapter
+ *	@q: the Tx queue to reclaim descriptors from
+ *	@n: the number of descriptors to reclaim
+ *	@unmap: whether the buffers should be unmapped for DMA
+ *
+ *	Reclaims Tx descriptors from an SGE Tx queue and frees the associated
+ *	Tx buffers.  Called with the Tx queue lock held.
+ */
+static void free_tx_desc(struct adapter *adap, struct sge_txq *q,
+			 unsigned int n, bool unmap)
+{
+	struct tx_sw_desc *d;
+	unsigned int cidx = q->cidx;
+	struct device *dev = adap->pdev_dev;
+
+	d = &q->sdesc[cidx];
+	while (n--) {
+		if (d->skb) {                       /* an SGL is present */
+			if (unmap)
+				unmap_sgl(dev, d->skb, d->sgl, q);
+			kfree_skb(d->skb);
+			d->skb = NULL;
+		}
+		++d;
+		if (++cidx == q->size) {
+			cidx = 0;
+			d = q->sdesc;
+		}
+	}
+	q->cidx = cidx;
+}
+
+/*
+ * Return the number of reclaimable descriptors in a Tx queue.
+ */
+static inline int reclaimable(const struct sge_txq *q)
+{
+	int hw_cidx = ntohs(q->stat->cidx);
+	hw_cidx -= q->cidx;
+	return hw_cidx < 0 ? hw_cidx + q->size : hw_cidx;
+}
+
+/**
+ *	reclaim_completed_tx - reclaims completed Tx descriptors
+ *	@adap: the adapter
+ *	@q: the Tx queue to reclaim completed descriptors from
+ *	@unmap: whether the buffers should be unmapped for DMA
+ *
+ *	Reclaims Tx descriptors that the SGE has indicated it has processed,
+ *	and frees the associated buffers if possible.  Called with the Tx
+ *	queue locked.
+ */
+static inline void reclaim_completed_tx(struct adapter *adap, struct sge_txq *q,
+					bool unmap)
+{
+	int avail = reclaimable(q);
+
+	if (avail) {
+		/*
+		 * Limit the amount of clean up work we do at a time to keep
+		 * the Tx lock hold time O(1).
+		 */
+		if (avail > MAX_TX_RECLAIM)
+			avail = MAX_TX_RECLAIM;
+
+		free_tx_desc(adap, q, avail, unmap);
+		q->in_use -= avail;
+	}
+}
+
+static inline int get_buf_size(const struct rx_sw_desc *d)
+{
+#if FL_PG_ORDER > 0
+	return (d->dma_addr & RX_LARGE_BUF) ? (PAGE_SIZE << FL_PG_ORDER) :
+					      PAGE_SIZE;
+#else
+	return PAGE_SIZE;
+#endif
+}
+
+/**
+ *	free_rx_bufs - free the Rx buffers on an SGE free list
+ *	@adap: the adapter
+ *	@q: the SGE free list to free buffers from
+ *	@n: how many buffers to free
+ *
+ *	Release the next @n buffers on an SGE free-buffer Rx queue.   The
+ *	buffers must be made inaccessible to HW before calling this function.
+ */
+static void free_rx_bufs(struct adapter *adap, struct sge_fl *q, int n)
+{
+	while (n--) {
+		struct rx_sw_desc *d = &q->sdesc[q->cidx];
+
+		if (is_buf_mapped(d))
+			dma_unmap_page(adap->pdev_dev, get_buf_addr(d),
+				       get_buf_size(d), PCI_DMA_FROMDEVICE);
+		put_page(d->page);
+		d->page = NULL;
+		if (++q->cidx == q->size)
+			q->cidx = 0;
+		q->avail--;
+	}
+}
+
+/**
+ *	unmap_rx_buf - unmap the current Rx buffer on an SGE free list
+ *	@adap: the adapter
+ *	@q: the SGE free list
+ *
+ *	Unmap the current buffer on an SGE free-buffer Rx queue.   The
+ *	buffer must be made inaccessible to HW before calling this function.
+ *
+ *	This is similar to @free_rx_bufs above but does not free the buffer.
+ *	Do note that the FL still loses any further access to the buffer.
+ */
+static void unmap_rx_buf(struct adapter *adap, struct sge_fl *q)
+{
+	struct rx_sw_desc *d = &q->sdesc[q->cidx];
+
+	if (is_buf_mapped(d))
+		dma_unmap_page(adap->pdev_dev, get_buf_addr(d),
+			       get_buf_size(d), PCI_DMA_FROMDEVICE);
+	d->page = NULL;
+	if (++q->cidx == q->size)
+		q->cidx = 0;
+	q->avail--;
+}
+
+static inline void ring_fl_db(struct adapter *adap, struct sge_fl *q)
+{
+	if (q->pend_cred >= 8) {
+		t4_write_reg(adap, MYPF_REG(SGE_PF_KDOORBELL), DBPRIO |
+			     QID(q->cntxt_id) | PIDX(q->pend_cred / 8));
+		q->pend_cred &= 7;
+	}
+}
+
+static inline void set_rx_sw_desc(struct rx_sw_desc *sd, struct page *pg,
+				  dma_addr_t mapping)
+{
+	sd->page = pg;
+	sd->dma_addr = mapping;      /* includes size low bits */
+}
+
+/**
+ *	refill_fl - refill an SGE Rx buffer ring
+ *	@adap: the adapter
+ *	@q: the ring to refill
+ *	@n: the number of new buffers to allocate
+ *	@gfp: the gfp flags for the allocations
+ *
+ *	(Re)populate an SGE free-buffer queue with up to @n new packet buffers,
+ *	allocated with the supplied gfp flags.  The caller must assure that
+ *	@n does not exceed the queue's capacity.  Returns the number of buffers
+ *	allocated.
+ */
+static unsigned int refill_fl(struct adapter *adap, struct sge_fl *q, int n,
+			      gfp_t gfp)
+{
+	struct page *pg;
+	dma_addr_t mapping;
+	unsigned int cred = q->avail;
+	__be64 *d = &q->desc[q->pidx];
+	struct rx_sw_desc *sd = &q->sdesc[q->pidx];
+
+	gfp |= __GFP_NOWARN;         /* failures are expected */
+
+#if FL_PG_ORDER > 0
+	/*
+	 * Prefer large buffers
+	 */
+	while (n) {
+		pg = alloc_pages(gfp | __GFP_COMP, FL_PG_ORDER);
+		if (unlikely(!pg)) {
+			q->large_alloc_failed++;
+			break;       /* fall back to single pages */
+		}
+
+		mapping = dma_map_page(adap->pdev_dev, pg, 0,
+				       PAGE_SIZE << FL_PG_ORDER,
+				       PCI_DMA_FROMDEVICE);
+		if (unlikely(dma_mapping_error(adap->pdev_dev, mapping))) {
+			__free_pages(pg, FL_PG_ORDER);
+			goto out;   /* do not try small pages for this error */
+		}
+		mapping |= RX_LARGE_BUF;
+		*d++ = cpu_to_be64(mapping);
+
+		set_rx_sw_desc(sd, pg, mapping);
+		sd++;
+
+		q->avail++;
+		if (++q->pidx == q->size) {
+			q->pidx = 0;
+			sd = q->sdesc;
+			d = q->desc;
+		}
+		n--;
+	}
+#endif
+
+	while (n--) {
+		pg = __netdev_alloc_page(adap->port[0], gfp);
+		if (unlikely(!pg)) {
+			q->alloc_failed++;
+			break;
+		}
+
+		mapping = dma_map_page(adap->pdev_dev, pg, 0, PAGE_SIZE,
+				       PCI_DMA_FROMDEVICE);
+		if (unlikely(dma_mapping_error(adap->pdev_dev, mapping))) {
+			netdev_free_page(adap->port[0], pg);
+			break;
+		}
+		*d++ = cpu_to_be64(mapping);
+
+		set_rx_sw_desc(sd, pg, mapping);
+		sd++;
+
+		q->avail++;
+		if (++q->pidx == q->size) {
+			q->pidx = 0;
+			sd = q->sdesc;
+			d = q->desc;
+		}
+	}
+
+out:	cred = q->avail - cred;
+	q->pend_cred += cred;
+	ring_fl_db(adap, q);
+	return cred;
+}
+
+static inline void __refill_fl(struct adapter *adap, struct sge_fl *fl)
+{
+	refill_fl(adap, fl, min(MAX_RX_REFILL, fl_cap(fl) - fl->avail),
+		  GFP_ATOMIC);
+}
+
+/**
+ *	replenish_fl - refill an SGE Rx buffer ring and check for starvation
+ *	@adap: the adapter
+ *	@fl: the ring to refill
+ *	@gfp: the gfp flags for the allocations
+ *
+ *	Attempt to refill an FL to capacity.  If afterwards the queue is found
+ *	critically low mark it as starving in the bitmap of starving FLs.
+ */
+static void replenish_fl(struct adapter *adap, struct sge_fl *fl, gfp_t gfp)
+{
+	refill_fl(adap, fl, fl_cap(fl) - fl->avail, gfp);
+	if (unlikely(fl_starving(fl)))
+		set_bit(fl->cntxt_id, adap->sge.starving_fl);
+}
+
+/**
+ *	alloc_ring - allocate resources for an SGE descriptor ring
+ *	@dev: the PCI device's core device
+ *	@nelem: the number of descriptors
+ *	@elem_size: the size of each descriptor
+ *	@sw_size: the size of the SW state associated with each ring element
+ *	@phys: the physical address of the allocated ring
+ *	@metadata: address of the array holding the SW state for the ring
+ *	@stat_size: extra space in HW ring for status information
+ *
+ *	Allocates resources for an SGE descriptor ring, such as Tx queues,
+ *	free buffer lists, or response queues.  Each SGE ring requires
+ *	space for its HW descriptors plus, optionally, space for the SW state
+ *	associated with each HW entry (the metadata).  The function returns
+ *	three values: the virtual address for the HW ring (the return value
+ *	of the function), the bus address of the HW ring, and the address
+ *	of the SW ring.
+ */
+static void *alloc_ring(struct device *dev, size_t nelem, size_t elem_size,
+			size_t sw_size, dma_addr_t *phys, void *metadata,
+			size_t stat_size)
+{
+	size_t len = nelem * elem_size + stat_size;
+	void *s = NULL;
+	void *p = dma_alloc_coherent(dev, len, phys, GFP_KERNEL);
+
+	if (!p)
+		return NULL;
+	if (sw_size) {
+		s = kcalloc(nelem, sw_size, GFP_KERNEL);
+
+		if (!s) {
+			dma_free_coherent(dev, len, p, *phys);
+			return NULL;
+		}
+	}
+	if (metadata)
+		*(void **)metadata = s;
+	memset(p, 0, len);
+	return p;
+}
+
+/**
+ *	sgl_len - calculates the size of an SGL of the given capacity
+ *	@n: the number of SGL entries
+ *
+ *	Calculates the number of flits needed for a scatter/gather list that
+ *	can hold the given number of entries.
+ */
+static inline unsigned int sgl_len(unsigned int n)
+{
+	n--;
+	return (3 * n) / 2 + (n & 1) + 2;
+}
+
+/**
+ *	flits_to_desc - returns the num of Tx descriptors for the given flits
+ *	@n: the number of flits
+ *
+ *	Returns the number of Tx descriptors needed for the supplied number
+ *	of flits.
+ */
+static inline unsigned int flits_to_desc(unsigned int n)
+{
+	BUG_ON(n > SGE_MAX_WR_LEN / 8);
+	return DIV_ROUND_UP(n, 8);
+}
+
+/**
+ *	is_eth_imm - can an Ethernet packet be sent as immediate data?
+ *	@skb: the packet
+ *
+ *	Returns whether an Ethernet packet is small enough to fit as
+ *	immediate data.
+ */
+static inline int is_eth_imm(const struct sk_buff *skb)
+{
+	return skb->len <= MAX_IMM_TX_PKT_LEN - sizeof(struct cpl_tx_pkt);
+}
+
+/**
+ *	calc_tx_flits - calculate the number of flits for a packet Tx WR
+ *	@skb: the packet
+ *
+ * 	Returns the number of flits needed for a Tx WR for the given Ethernet
+ * 	packet, including the needed WR and CPL headers.
+ */
+static inline unsigned int calc_tx_flits(const struct sk_buff *skb)
+{
+	unsigned int flits;
+
+	if (is_eth_imm(skb))
+		return DIV_ROUND_UP(skb->len + sizeof(struct cpl_tx_pkt), 8);
+
+	flits = sgl_len(skb_shinfo(skb)->nr_frags + 1) + 4;
+	if (skb_shinfo(skb)->gso_size)
+		flits += 2;
+	return flits;
+}
+
+/**
+ *	calc_tx_descs - calculate the number of Tx descriptors for a packet
+ *	@skb: the packet
+ *
+ * 	Returns the number of Tx descriptors needed for the given Ethernet
+ * 	packet, including the needed WR and CPL headers.
+ */
+static inline unsigned int calc_tx_descs(const struct sk_buff *skb)
+{
+	return flits_to_desc(calc_tx_flits(skb));
+}
+
+/**
+ *	write_sgl - populate a scatter/gather list for a packet
+ *	@skb: the packet
+ *	@q: the Tx queue we are writing into
+ *	@sgl: starting location for writing the SGL
+ *	@end: points right after the end of the SGL
+ *	@start: start offset into skb main-body data to include in the SGL
+ *	@addr: the list of bus addresses for the SGL elements
+ *
+ *	Generates a gather list for the buffers that make up a packet.
+ *	The caller must provide adequate space for the SGL that will be written.
+ *	The SGL includes all of the packet's page fragments and the data in its
+ *	main body except for the first @start bytes.  @sgl must be 16-byte
+ *	aligned and within a Tx descriptor with available space.  @end points
+ *	right after the end of the SGL but does not account for any potential
+ *	wrap around, i.e., @end > @sgl.
+ */
+static void write_sgl(const struct sk_buff *skb, struct sge_txq *q,
+		      struct ulptx_sgl *sgl, u64 *end, unsigned int start,
+		      const dma_addr_t *addr)
+{
+	unsigned int i, len;
+	struct ulptx_sge_pair *to;
+	const struct skb_shared_info *si = skb_shinfo(skb);
+	unsigned int nfrags = si->nr_frags;
+	struct ulptx_sge_pair buf[MAX_SKB_FRAGS / 2 + 1];
+
+	len = skb_headlen(skb) - start;
+	if (likely(len)) {
+		sgl->len0 = htonl(len);
+		sgl->addr0 = cpu_to_be64(addr[0] + start);
+		nfrags++;
+	} else {
+		sgl->len0 = htonl(si->frags[0].size);
+		sgl->addr0 = cpu_to_be64(addr[1]);
+	}
+
+	sgl->cmd_nsge = htonl(ULPTX_CMD(ULP_TX_SC_DSGL) | ULPTX_NSGE(nfrags));
+	if (likely(--nfrags == 0))
+		return;
+	/*
+	 * Most of the complexity below deals with the possibility we hit the
+	 * end of the queue in the middle of writing the SGL.  For this case
+	 * only we create the SGL in a temporary buffer and then copy it.
+	 */
+	to = (u8 *)end > (u8 *)q->stat ? buf : sgl->sge;
+
+	for (i = (nfrags != si->nr_frags); nfrags >= 2; nfrags -= 2, to++) {
+		to->len[0] = cpu_to_be32(si->frags[i].size);
+		to->len[1] = cpu_to_be32(si->frags[++i].size);
+		to->addr[0] = cpu_to_be64(addr[i]);
+		to->addr[1] = cpu_to_be64(addr[++i]);
+	}
+	if (nfrags) {
+		to->len[0] = cpu_to_be32(si->frags[i].size);
+		to->len[1] = cpu_to_be32(0);
+		to->addr[0] = cpu_to_be64(addr[i + 1]);
+	}
+	if (unlikely((u8 *)end > (u8 *)q->stat)) {
+		unsigned int part0 = (u8 *)q->stat - (u8 *)sgl->sge, part1;
+
+		if (likely(part0))
+			memcpy(sgl->sge, buf, part0);
+		part1 = (u8 *)end - (u8 *)q->stat;
+		memcpy(q->desc, (u8 *)buf + part0, part1);
+		end = (void *)q->desc + part1;
+	}
+	if ((uintptr_t)end & 8)           /* 0-pad to multiple of 16 */
+		*(u64 *)end = 0;
+}
+
+/**
+ *	ring_tx_db - check and potentially ring a Tx queue's doorbell
+ *	@adap: the adapter
+ *	@q: the Tx queue
+ *	@n: number of new descriptors to give to HW
+ *
+ *	Ring the doorbel for a Tx queue.
+ */
+static inline void ring_tx_db(struct adapter *adap, struct sge_txq *q, int n)
+{
+	wmb();            /* write descriptors before telling HW */
+	t4_write_reg(adap, MYPF_REG(SGE_PF_KDOORBELL),
+		     QID(q->cntxt_id) | PIDX(n));
+}
+
+/**
+ * 	inline_tx_skb - inline a packet's data into Tx descriptors
+ * 	@skb: the packet
+ * 	@q: the Tx queue where the packet will be inlined
+ * 	@pos: starting position in the Tx queue where to inline the packet
+ *
+ *	Inline a packet's contents directly into Tx descriptors, starting at
+ *	the given position within the Tx DMA ring.
+ *	Most of the complexity of this operation is dealing with wrap arounds
+ *	in the middle of the packet we want to inline.
+ */
+static void inline_tx_skb(const struct sk_buff *skb, const struct sge_txq *q,
+			  void *pos)
+{
+	u64 *p;
+	int left = (void *)q->stat - pos;
+
+	if (likely(skb->len <= left)) {
+		if (likely(!skb->data_len))
+			skb_copy_from_linear_data(skb, pos, skb->len);
+		else
+			skb_copy_bits(skb, 0, pos, skb->len);
+		pos += skb->len;
+	} else {
+		skb_copy_bits(skb, 0, pos, left);
+		skb_copy_bits(skb, left, q->desc, skb->len - left);
+		pos = (void *)q->desc + (skb->len - left);
+	}
+
+	/* 0-pad to multiple of 16 */
+	p = PTR_ALIGN(pos, 8);
+	if ((uintptr_t)p & 8)
+		*p = 0;
+}
+
+/*
+ * Figure out what HW csum a packet wants and return the appropriate control
+ * bits.
+ */
+static u64 hwcsum(const struct sk_buff *skb)
+{
+	int csum_type;
+	const struct iphdr *iph = ip_hdr(skb);
+
+	if (iph->version == 4) {
+		if (iph->protocol == IPPROTO_TCP)
+			csum_type = TX_CSUM_TCPIP;
+		else if (iph->protocol == IPPROTO_UDP)
+			csum_type = TX_CSUM_UDPIP;
+		else {
+nocsum:			/*
+			 * unknown protocol, disable HW csum
+			 * and hope a bad packet is detected
+			 */
+			return TXPKT_L4CSUM_DIS;
+		}
+	} else {
+		/*
+		 * this doesn't work with extension headers
+		 */
+		const struct ipv6hdr *ip6h = (const struct ipv6hdr *)iph;
+
+		if (ip6h->nexthdr == IPPROTO_TCP)
+			csum_type = TX_CSUM_TCPIP6;
+		else if (ip6h->nexthdr == IPPROTO_UDP)
+			csum_type = TX_CSUM_UDPIP6;
+		else
+			goto nocsum;
+	}
+
+	if (likely(csum_type >= TX_CSUM_TCPIP))
+		return TXPKT_CSUM_TYPE(csum_type) |
+			TXPKT_IPHDR_LEN(skb_network_header_len(skb)) |
+			TXPKT_ETHHDR_LEN(skb_network_offset(skb) - ETH_HLEN);
+	else {
+		int start = skb_transport_offset(skb);
+
+		return TXPKT_CSUM_TYPE(csum_type) | TXPKT_CSUM_START(start) |
+			TXPKT_CSUM_LOC(start + skb->csum_offset);
+	}
+}
+
+static void eth_txq_stop(struct sge_eth_txq *q)
+{
+	netif_tx_stop_queue(q->txq);
+	q->q.stops++;
+}
+
+static inline void txq_advance(struct sge_txq *q, unsigned int n)
+{
+	q->in_use += n;
+	q->pidx += n;
+	if (q->pidx >= q->size)
+		q->pidx -= q->size;
+}
+
+/**
+ *	t4_eth_xmit - add a packet to an Ethernet Tx queue
+ *	@skb: the packet
+ *	@dev: the egress net device
+ *
+ *	Add a packet to an SGE Ethernet Tx queue.  Runs with softirqs disabled.
+ */
+netdev_tx_t t4_eth_xmit(struct sk_buff *skb, struct net_device *dev)
+{
+	u32 wr_mid;
+	u64 cntrl, *end;
+	int qidx, credits;
+	unsigned int flits, ndesc;
+	struct adapter *adap;
+	struct sge_eth_txq *q;
+	const struct port_info *pi;
+	struct fw_eth_tx_pkt_wr *wr;
+	struct cpl_tx_pkt_core *cpl;
+	const struct skb_shared_info *ssi;
+	dma_addr_t addr[MAX_SKB_FRAGS + 1];
+
+	/*
+	 * The chip min packet length is 10 octets but play safe and reject
+	 * anything shorter than an Ethernet header.
+	 */
+	if (unlikely(skb->len < ETH_HLEN)) {
+out_free:	dev_kfree_skb(skb);
+		return NETDEV_TX_OK;
+	}
+
+	pi = netdev_priv(dev);
+	adap = pi->adapter;
+	qidx = skb_get_queue_mapping(skb);
+	q = &adap->sge.ethtxq[qidx + pi->first_qset];
+
+	reclaim_completed_tx(adap, &q->q, true);
+
+	flits = calc_tx_flits(skb);
+	ndesc = flits_to_desc(flits);
+	credits = txq_avail(&q->q) - ndesc;
+
+	if (unlikely(credits < 0)) {
+		eth_txq_stop(q);
+		dev_err(adap->pdev_dev,
+			"%s: Tx ring %u full while queue awake!\n",
+			dev->name, qidx);
+		return NETDEV_TX_BUSY;
+	}
+
+	if (!is_eth_imm(skb) &&
+	    unlikely(map_skb(adap->pdev_dev, skb, addr) < 0)) {
+		q->mapping_err++;
+		goto out_free;
+	}
+
+	wr_mid = FW_WR_LEN16(DIV_ROUND_UP(flits, 2));
+	if (unlikely(credits < ETHTXQ_STOP_THRES)) {
+		eth_txq_stop(q);
+		wr_mid |= FW_WR_EQUEQ | FW_WR_EQUIQ;
+	}
+
+	wr = (void *)&q->q.desc[q->q.pidx];
+	wr->equiq_to_len16 = htonl(wr_mid);
+	wr->r3 = cpu_to_be64(0);
+	end = (u64 *)wr + flits;
+
+	ssi = skb_shinfo(skb);
+	if (ssi->gso_size) {
+		struct cpl_tx_pkt_lso *lso = (void *)wr;
+		bool v6 = (ssi->gso_type & SKB_GSO_TCPV6) != 0;
+		int l3hdr_len = skb_network_header_len(skb);
+		int eth_xtra_len = skb_network_offset(skb) - ETH_HLEN;
+
+		wr->op_immdlen = htonl(FW_WR_OP(FW_ETH_TX_PKT_WR) |
+				       FW_WR_IMMDLEN(sizeof(*lso)));
+		lso->lso_ctrl = htonl(LSO_OPCODE(CPL_TX_PKT_LSO) |
+				      LSO_FIRST_SLICE | LSO_LAST_SLICE |
+				      LSO_IPV6(v6) |
+				      LSO_ETHHDR_LEN(eth_xtra_len / 4) |
+				      LSO_IPHDR_LEN(l3hdr_len / 4) |
+				      LSO_TCPHDR_LEN(tcp_hdr(skb)->doff));
+		lso->ipid_ofst = htons(0);
+		lso->mss = htons(ssi->gso_size);
+		lso->seqno_offset = htonl(0);
+		lso->len = htonl(skb->len);
+		cpl = (void *)(lso + 1);
+		cntrl = TXPKT_CSUM_TYPE(v6 ? TX_CSUM_TCPIP6 : TX_CSUM_TCPIP) |
+			TXPKT_IPHDR_LEN(l3hdr_len) |
+			TXPKT_ETHHDR_LEN(eth_xtra_len);
+		q->tso++;
+		q->tx_cso += ssi->gso_segs;
+	} else {
+		int len;
+
+		len = is_eth_imm(skb) ? skb->len + sizeof(*cpl) : sizeof(*cpl);
+		wr->op_immdlen = htonl(FW_WR_OP(FW_ETH_TX_PKT_WR) |
+				       FW_WR_IMMDLEN(len));
+		cpl = (void *)(wr + 1);
+		if (skb->ip_summed == CHECKSUM_PARTIAL) {
+			cntrl = hwcsum(skb) | TXPKT_IPCSUM_DIS;
+			q->tx_cso++;
+		} else
+			cntrl = TXPKT_L4CSUM_DIS | TXPKT_IPCSUM_DIS;
+	}
+
+	if (vlan_tx_tag_present(skb)) {
+		q->vlan_ins++;
+		cntrl |= TXPKT_VLAN_VLD | TXPKT_VLAN(vlan_tx_tag_get(skb));
+	}
+
+	cpl->ctrl0 = htonl(TXPKT_OPCODE(CPL_TX_PKT_XT) |
+			   TXPKT_INTF(pi->tx_chan) | TXPKT_PF(0));
+	cpl->pack = htons(0);
+	cpl->len = htons(skb->len);
+	cpl->ctrl1 = cpu_to_be64(cntrl);
+
+	if (is_eth_imm(skb)) {
+		inline_tx_skb(skb, &q->q, cpl + 1);
+		dev_kfree_skb(skb);
+	} else {
+		int last_desc;
+
+		write_sgl(skb, &q->q, (struct ulptx_sgl *)(cpl + 1), end, 0,
+			  addr);
+		skb_orphan(skb);
+
+		last_desc = q->q.pidx + ndesc - 1;
+		if (last_desc >= q->q.size)
+			last_desc -= q->q.size;
+		q->q.sdesc[last_desc].skb = skb;
+		q->q.sdesc[last_desc].sgl = (struct ulptx_sgl *)(cpl + 1);
+	}
+
+	txq_advance(&q->q, ndesc);
+
+	ring_tx_db(adap, &q->q, ndesc);
+	return NETDEV_TX_OK;
+}
+
+/**
+ *	reclaim_completed_tx_imm - reclaim completed control-queue Tx descs
+ *	@q: the SGE control Tx queue
+ *
+ *	This is a variant of reclaim_completed_tx() that is used for Tx queues
+ *	that send only immediate data (presently just the control queues) and
+ *	thus do not have any sk_buffs to release.
+ */
+static inline void reclaim_completed_tx_imm(struct sge_txq *q)
+{
+	int hw_cidx = ntohs(q->stat->cidx);
+	int reclaim = hw_cidx - q->cidx;
+
+	if (reclaim < 0)
+		reclaim += q->size;
+
+	q->in_use -= reclaim;
+	q->cidx = hw_cidx;
+}
+
+/**
+ *	is_imm - check whether a packet can be sent as immediate data
+ *	@skb: the packet
+ *
+ *	Returns true if a packet can be sent as a WR with immediate data.
+ */
+static inline int is_imm(const struct sk_buff *skb)
+{
+	return skb->len <= MAX_CTRL_WR_LEN;
+}
+
+/**
+ *	ctrlq_check_stop - check if a control queue is full and should stop
+ *	@q: the queue
+ *	@wr: most recent WR written to the queue
+ *
+ *	Check if a control queue has become full and should be stopped.
+ *	We clean up control queue descriptors very lazily, only when we are out.
+ *	If the queue is still full after reclaiming any completed descriptors
+ *	we suspend it and have the last WR wake it up.
+ */
+static void ctrlq_check_stop(struct sge_ctrl_txq *q, struct fw_wr_hdr *wr)
+{
+	reclaim_completed_tx_imm(&q->q);
+	if (unlikely(txq_avail(&q->q) < TXQ_STOP_THRES)) {
+		wr->lo |= htonl(FW_WR_EQUEQ | FW_WR_EQUIQ);
+		q->q.stops++;
+		q->full = 1;
+	}
+}
+
+/**
+ *	ctrl_xmit - send a packet through an SGE control Tx queue
+ *	@q: the control queue
+ *	@skb: the packet
+ *
+ *	Send a packet through an SGE control Tx queue.  Packets sent through
+ *	a control queue must fit entirely as immediate data.
+ */
+static int ctrl_xmit(struct sge_ctrl_txq *q, struct sk_buff *skb)
+{
+	unsigned int ndesc;
+	struct fw_wr_hdr *wr;
+
+	if (unlikely(!is_imm(skb))) {
+		WARN_ON(1);
+		dev_kfree_skb(skb);
+		return NET_XMIT_DROP;
+	}
+
+	ndesc = DIV_ROUND_UP(skb->len, sizeof(struct tx_desc));
+	spin_lock(&q->sendq.lock);
+
+	if (unlikely(q->full)) {
+		skb->priority = ndesc;                  /* save for restart */
+		__skb_queue_tail(&q->sendq, skb);
+		spin_unlock(&q->sendq.lock);
+		return NET_XMIT_CN;
+	}
+
+	wr = (struct fw_wr_hdr *)&q->q.desc[q->q.pidx];
+	inline_tx_skb(skb, &q->q, wr);
+
+	txq_advance(&q->q, ndesc);
+	if (unlikely(txq_avail(&q->q) < TXQ_STOP_THRES))
+		ctrlq_check_stop(q, wr);
+
+	ring_tx_db(q->adap, &q->q, ndesc);
+	spin_unlock(&q->sendq.lock);
+
+	kfree_skb(skb);
+	return NET_XMIT_SUCCESS;
+}
+
+/**
+ *	restart_ctrlq - restart a suspended control queue
+ *	@data: the control queue to restart
+ *
+ *	Resumes transmission on a suspended Tx control queue.
+ */
+static void restart_ctrlq(unsigned long data)
+{
+	struct sk_buff *skb;
+	unsigned int written = 0;
+	struct sge_ctrl_txq *q = (struct sge_ctrl_txq *)data;
+
+	spin_lock(&q->sendq.lock);
+	reclaim_completed_tx_imm(&q->q);
+	BUG_ON(txq_avail(&q->q) < TXQ_STOP_THRES);  /* q should be empty */
+
+	while ((skb = __skb_dequeue(&q->sendq)) != NULL) {
+		struct fw_wr_hdr *wr;
+		unsigned int ndesc = skb->priority;     /* previously saved */
+
+		/*
+		 * Write descriptors and free skbs outside the lock to limit
+		 * wait times.  q->full is still set so new skbs will be queued.
+		 */
+		spin_unlock(&q->sendq.lock);
+
+		wr = (struct fw_wr_hdr *)&q->q.desc[q->q.pidx];
+		inline_tx_skb(skb, &q->q, wr);
+		kfree_skb(skb);
+
+		written += ndesc;
+		txq_advance(&q->q, ndesc);
+		if (unlikely(txq_avail(&q->q) < TXQ_STOP_THRES)) {
+			unsigned long old = q->q.stops;
+
+			ctrlq_check_stop(q, wr);
+			if (q->q.stops != old) {          /* suspended anew */
+				spin_lock(&q->sendq.lock);
+				goto ringdb;
+			}
+		}
+		if (written > 16) {
+			ring_tx_db(q->adap, &q->q, written);
+			written = 0;
+		}
+		spin_lock(&q->sendq.lock);
+	}
+	q->full = 0;
+ringdb: if (written)
+		ring_tx_db(q->adap, &q->q, written);
+	spin_unlock(&q->sendq.lock);
+}
+
+/**
+ *	t4_mgmt_tx - send a management message
+ *	@adap: the adapter
+ *	@skb: the packet containing the management message
+ *
+ *	Send a management message through control queue 0.
+ */
+int t4_mgmt_tx(struct adapter *adap, struct sk_buff *skb)
+{
+	int ret;
+
+	local_bh_disable();
+	ret = ctrl_xmit(&adap->sge.ctrlq[0], skb);
+	local_bh_enable();
+	return ret;
+}
+
+/**
+ *	deferred_unmap_destructor - unmap a packet when it is freed
+ *	@skb: the packet
+ *
+ *	This is the packet destructor used for Tx packets that need to remain
+ *	mapped until they are freed rather than until their Tx descriptors are
+ *	freed.
+ */
+static void deferred_unmap_destructor(struct sk_buff *skb)
+{
+	unmap_skb(skb->dev->dev.parent, skb, (dma_addr_t *)skb->head);
+}
+
+/**
+ *	is_ofld_imm - check whether a packet can be sent as immediate data
+ *	@skb: the packet
+ *
+ *	Returns true if a packet can be sent as an offload WR with immediate
+ *	data.  We currently use the same limit as for Ethernet packets.
+ */
+static inline int is_ofld_imm(const struct sk_buff *skb)
+{
+	return skb->len <= MAX_IMM_TX_PKT_LEN;
+}
+
+/**
+ *	calc_tx_flits_ofld - calculate # of flits for an offload packet
+ *	@skb: the packet
+ *
+ * 	Returns the number of flits needed for the given offload packet.
+ * 	These packets are already fully constructed and no additional headers
+ * 	will be added.
+ */
+static inline unsigned int calc_tx_flits_ofld(const struct sk_buff *skb)
+{
+	unsigned int flits, cnt;
+
+	if (is_ofld_imm(skb))
+		return DIV_ROUND_UP(skb->len, 8);
+
+	flits = skb_transport_offset(skb) / 8U;   /* headers */
+	cnt = skb_shinfo(skb)->nr_frags;
+	if (skb->tail != skb->transport_header)
+		cnt++;
+	return flits + sgl_len(cnt);
+}
+
+/**
+ *	txq_stop_maperr - stop a Tx queue due to I/O MMU exhaustion
+ *	@adap: the adapter
+ *	@q: the queue to stop
+ *
+ *	Mark a Tx queue stopped due to I/O MMU exhaustion and resulting
+ *	inability to map packets.  A periodic timer attempts to restart
+ *	queues so marked.
+ */
+static void txq_stop_maperr(struct sge_ofld_txq *q)
+{
+	q->mapping_err++;
+	q->q.stops++;
+	set_bit(q->q.cntxt_id, q->adap->sge.txq_maperr);
+}
+
+/**
+ *	ofldtxq_stop - stop an offload Tx queue that has become full
+ *	@q: the queue to stop
+ *	@skb: the packet causing the queue to become full
+ *
+ *	Stops an offload Tx queue that has become full and modifies the packet
+ *	being written to request a wakeup.
+ */
+static void ofldtxq_stop(struct sge_ofld_txq *q, struct sk_buff *skb)
+{
+	struct fw_wr_hdr *wr = (struct fw_wr_hdr *)skb->data;
+
+	wr->lo |= htonl(FW_WR_EQUEQ | FW_WR_EQUIQ);
+	q->q.stops++;
+	q->full = 1;
+}
+
+/**
+ *	service_ofldq - restart a suspended offload queue
+ *	@q: the offload queue
+ *
+ *	Services an offload Tx queue by moving packets from its packet queue
+ *	to the HW Tx ring.  The function starts and ends with the queue locked.
+ */
+static void service_ofldq(struct sge_ofld_txq *q)
+{
+	u64 *pos;
+	int credits;
+	struct sk_buff *skb;
+	unsigned int written = 0;
+	unsigned int flits, ndesc;
+
+	while ((skb = skb_peek(&q->sendq)) != NULL && !q->full) {
+		/*
+		 * We drop the lock but leave skb on sendq, thus retaining
+		 * exclusive access to the state of the queue.
+		 */
+		spin_unlock(&q->sendq.lock);
+
+		reclaim_completed_tx(q->adap, &q->q, false);
+
+		flits = skb->priority;                /* previously saved */
+		ndesc = flits_to_desc(flits);
+		credits = txq_avail(&q->q) - ndesc;
+		BUG_ON(credits < 0);
+		if (unlikely(credits < TXQ_STOP_THRES))
+			ofldtxq_stop(q, skb);
+
+		pos = (u64 *)&q->q.desc[q->q.pidx];
+		if (is_ofld_imm(skb))
+			inline_tx_skb(skb, &q->q, pos);
+		else if (map_skb(q->adap->pdev_dev, skb,
+				 (dma_addr_t *)skb->head)) {
+			txq_stop_maperr(q);
+			spin_lock(&q->sendq.lock);
+			break;
+		} else {
+			int last_desc, hdr_len = skb_transport_offset(skb);
+
+			memcpy(pos, skb->data, hdr_len);
+			write_sgl(skb, &q->q, (void *)pos + hdr_len,
+				  pos + flits, hdr_len,
+				  (dma_addr_t *)skb->head);
+
+			skb->dev = q->adap->port[0];
+			skb->destructor = deferred_unmap_destructor;
+
+			last_desc = q->q.pidx + ndesc - 1;
+			if (last_desc >= q->q.size)
+				last_desc -= q->q.size;
+			q->q.sdesc[last_desc].skb = skb;
+		}
+
+		txq_advance(&q->q, ndesc);
+		written += ndesc;
+		if (unlikely(written > 32)) {
+			ring_tx_db(q->adap, &q->q, written);
+			written = 0;
+		}
+
+		spin_lock(&q->sendq.lock);
+		__skb_unlink(skb, &q->sendq);
+		if (is_ofld_imm(skb))
+			kfree_skb(skb);
+	}
+	if (likely(written))
+		ring_tx_db(q->adap, &q->q, written);
+}
+
+/**
+ *	ofld_xmit - send a packet through an offload queue
+ *	@q: the Tx offload queue
+ *	@skb: the packet
+ *
+ *	Send an offload packet through an SGE offload queue.
+ */
+static int ofld_xmit(struct sge_ofld_txq *q, struct sk_buff *skb)
+{
+	skb->priority = calc_tx_flits_ofld(skb);       /* save for restart */
+	spin_lock(&q->sendq.lock);
+	__skb_queue_tail(&q->sendq, skb);
+	if (q->sendq.qlen == 1)
+		service_ofldq(q);
+	spin_unlock(&q->sendq.lock);
+	return NET_XMIT_SUCCESS;
+}
+
+/**
+ *	restart_ofldq - restart a suspended offload queue
+ *	@data: the offload queue to restart
+ *
+ *	Resumes transmission on a suspended Tx offload queue.
+ */
+static void restart_ofldq(unsigned long data)
+{
+	struct sge_ofld_txq *q = (struct sge_ofld_txq *)data;
+
+	spin_lock(&q->sendq.lock);
+	q->full = 0;            /* the queue actually is completely empty now */
+	service_ofldq(q);
+	spin_unlock(&q->sendq.lock);
+}
+
+/**
+ *	skb_txq - return the Tx queue an offload packet should use
+ *	@skb: the packet
+ *
+ *	Returns the Tx queue an offload packet should use as indicated by bits
+ *	1-15 in the packet's queue_mapping.
+ */
+static inline unsigned int skb_txq(const struct sk_buff *skb)
+{
+	return skb->queue_mapping >> 1;
+}
+
+/**
+ *	is_ctrl_pkt - return whether an offload packet is a control packet
+ *	@skb: the packet
+ *
+ *	Returns whether an offload packet should use an OFLD or a CTRL
+ *	Tx queue as indicated by bit 0 in the packet's queue_mapping.
+ */
+static inline unsigned int is_ctrl_pkt(const struct sk_buff *skb)
+{
+	return skb->queue_mapping & 1;
+}
+
+static inline int ofld_send(struct adapter *adap, struct sk_buff *skb)
+{
+	unsigned int idx = skb_txq(skb);
+
+	if (unlikely(is_ctrl_pkt(skb)))
+		return ctrl_xmit(&adap->sge.ctrlq[idx], skb);
+	return ofld_xmit(&adap->sge.ofldtxq[idx], skb);
+}
+
+/**
+ *	t4_ofld_send - send an offload packet
+ *	@adap: the adapter
+ *	@skb: the packet
+ *
+ *	Sends an offload packet.  We use the packet queue_mapping to select the
+ *	appropriate Tx queue as follows: bit 0 indicates whether the packet
+ *	should be sent as regular or control, bits 1-15 select the queue.
+ */
+int t4_ofld_send(struct adapter *adap, struct sk_buff *skb)
+{
+	int ret;
+
+	local_bh_disable();
+	ret = ofld_send(adap, skb);
+	local_bh_enable();
+	return ret;
+}
+
+/**
+ *	cxgb4_ofld_send - send an offload packet
+ *	@dev: the net device
+ *	@skb: the packet
+ *
+ *	Sends an offload packet.  This is an exported version of @t4_ofld_send,
+ *	intended for ULDs.
+ */
+int cxgb4_ofld_send(struct net_device *dev, struct sk_buff *skb)
+{
+	return t4_ofld_send(netdev2adap(dev), skb);
+}
+EXPORT_SYMBOL(cxgb4_ofld_send);
+
+static inline void copy_frags(struct skb_shared_info *ssi,
+			      const struct pkt_gl *gl, unsigned int offset)
+{
+	unsigned int n;
+
+	/* usually there's just one frag */
+	ssi->frags[0].page = gl->frags[0].page;
+	ssi->frags[0].page_offset = gl->frags[0].page_offset + offset;
+	ssi->frags[0].size = gl->frags[0].size - offset;
+	ssi->nr_frags = gl->nfrags;
+	n = gl->nfrags - 1;
+	if (n)
+		memcpy(&ssi->frags[1], &gl->frags[1], n * sizeof(skb_frag_t));
+
+	/* get a reference to the last page, we don't own it */
+	get_page(gl->frags[n].page);
+}
+
+/**
+ *	cxgb4_pktgl_to_skb - build an sk_buff from a packet gather list
+ *	@gl: the gather list
+ *	@skb_len: size of sk_buff main body if it carries fragments
+ *	@pull_len: amount of data to move to the sk_buff's main body
+ *
+ *	Builds an sk_buff from the given packet gather list.  Returns the
+ *	sk_buff or %NULL if sk_buff allocation failed.
+ */
+struct sk_buff *cxgb4_pktgl_to_skb(const struct pkt_gl *gl,
+				   unsigned int skb_len, unsigned int pull_len)
+{
+	struct sk_buff *skb;
+
+	/*
+	 * Below we rely on RX_COPY_THRES being less than the smallest Rx buffer
+	 * size, which is expected since buffers are at least PAGE_SIZEd.
+	 * In this case packets up to RX_COPY_THRES have only one fragment.
+	 */
+	if (gl->tot_len <= RX_COPY_THRES) {
+		skb = dev_alloc_skb(gl->tot_len);
+		if (unlikely(!skb))
+			goto out;
+		__skb_put(skb, gl->tot_len);
+		skb_copy_to_linear_data(skb, gl->va, gl->tot_len);
+	} else {
+		skb = dev_alloc_skb(skb_len);
+		if (unlikely(!skb))
+			goto out;
+		__skb_put(skb, pull_len);
+		skb_copy_to_linear_data(skb, gl->va, pull_len);
+
+		copy_frags(skb_shinfo(skb), gl, pull_len);
+		skb->len = gl->tot_len;
+		skb->data_len = skb->len - pull_len;
+		skb->truesize += skb->data_len;
+	}
+out:	return skb;
+}
+EXPORT_SYMBOL(cxgb4_pktgl_to_skb);
+
+/**
+ *	t4_pktgl_free - free a packet gather list
+ *	@gl: the gather list
+ *
+ *	Releases the pages of a packet gather list.  We do not own the last
+ *	page on the list and do not free it.
+ */
+void t4_pktgl_free(const struct pkt_gl *gl)
+{
+	int n;
+	const skb_frag_t *p;
+
+	for (p = gl->frags, n = gl->nfrags - 1; n--; p++)
+		put_page(p->page);
+}
+
+/*
+ * Process an MPS trace packet.  Give it an unused protocol number so it won't
+ * be delivered to anyone and send it to the stack for capture.
+ */
+static noinline int handle_trace_pkt(struct adapter *adap,
+				     const struct pkt_gl *gl)
+{
+	struct sk_buff *skb;
+	struct cpl_trace_pkt *p;
+
+	skb = cxgb4_pktgl_to_skb(gl, RX_PULL_LEN, RX_PULL_LEN);
+	if (unlikely(!skb)) {
+		t4_pktgl_free(gl);
+		return 0;
+	}
+
+	p = (struct cpl_trace_pkt *)skb->data;
+	__skb_pull(skb, sizeof(*p));
+	skb_reset_mac_header(skb);
+	skb->protocol = htons(0xffff);
+	skb->dev = adap->port[0];
+	netif_receive_skb(skb);
+	return 0;
+}
+
+static void do_gro(struct sge_eth_rxq *rxq, const struct pkt_gl *gl,
+		   const struct cpl_rx_pkt *pkt)
+{
+	int ret;
+	struct sk_buff *skb;
+
+	skb = napi_get_frags(&rxq->rspq.napi);
+	if (unlikely(!skb)) {
+		t4_pktgl_free(gl);
+		rxq->stats.rx_drops++;
+		return;
+	}
+
+	copy_frags(skb_shinfo(skb), gl, RX_PKT_PAD);
+	skb->len = gl->tot_len - RX_PKT_PAD;
+	skb->data_len = skb->len;
+	skb->truesize += skb->data_len;
+	skb->ip_summed = CHECKSUM_UNNECESSARY;
+	skb_record_rx_queue(skb, rxq->rspq.idx);
+
+	if (unlikely(pkt->vlan_ex)) {
+		struct port_info *pi = netdev_priv(rxq->rspq.netdev);
+		struct vlan_group *grp = pi->vlan_grp;
+
+		rxq->stats.vlan_ex++;
+		if (likely(grp)) {
+			ret = vlan_gro_frags(&rxq->rspq.napi, grp,
+					     ntohs(pkt->vlan));
+			goto stats;
+		}
+	}
+	ret = napi_gro_frags(&rxq->rspq.napi);
+stats:	if (ret == GRO_HELD)
+		rxq->stats.lro_pkts++;
+	else if (ret == GRO_MERGED || ret == GRO_MERGED_FREE)
+		rxq->stats.lro_merged++;
+	rxq->stats.pkts++;
+	rxq->stats.rx_cso++;
+}
+
+/**
+ *	t4_ethrx_handler - process an ingress ethernet packet
+ *	@q: the response queue that received the packet
+ *	@rsp: the response queue descriptor holding the RX_PKT message
+ *	@si: the gather list of packet fragments
+ *
+ *	Process an ingress ethernet packet and deliver it to the stack.
+ */
+int t4_ethrx_handler(struct sge_rspq *q, const __be64 *rsp,
+		     const struct pkt_gl *si)
+{
+	bool csum_ok;
+	struct sk_buff *skb;
+	struct port_info *pi;
+	const struct cpl_rx_pkt *pkt;
+	struct sge_eth_rxq *rxq = container_of(q, struct sge_eth_rxq, rspq);
+
+	if (unlikely(*(u8 *)rsp == CPL_TRACE_PKT))
+		return handle_trace_pkt(q->adap, si);
+
+	pkt = (void *)&rsp[1];
+	csum_ok = pkt->csum_calc && !pkt->err_vec;
+	if ((pkt->l2info & htonl(RXF_TCP)) &&
+	    (q->netdev->features & NETIF_F_GRO) && csum_ok && !pkt->ip_frag) {
+		do_gro(rxq, si, pkt);
+		return 0;
+	}
+
+	if (si->tot_len <= RX_COPY_THRES) {
+		/* small packets have only one fragment */
+		skb = dev_alloc_skb(si->frags[0].size);
+		if (!skb)
+			goto nomem;
+		__skb_put(skb, si->frags[0].size);
+		skb_copy_to_linear_data(skb, si->va, si->frags[0].size);
+	} else {
+		skb = dev_alloc_skb(RX_PKT_SKB_LEN);
+		if (!skb)
+			goto nomem;
+		__skb_put(skb, RX_PULL_LEN);
+		skb_copy_to_linear_data(skb, si->va, RX_PULL_LEN);
+
+		copy_frags(skb_shinfo(skb), si, RX_PULL_LEN);
+		skb->len = si->tot_len;
+		skb->data_len = skb->len - RX_PULL_LEN;
+		skb->truesize += skb->data_len;
+	}
+
+	__skb_pull(skb, RX_PKT_PAD);      /* remove ethernet header padding */
+	skb->protocol = eth_type_trans(skb, q->netdev);
+	skb_record_rx_queue(skb, q->idx);
+	pi = netdev_priv(skb->dev);
+	rxq->stats.pkts++;
+
+	if (csum_ok && (pi->rx_offload & RX_CSO) &&
+	    (pkt->l2info & htonl(RXF_UDP | RXF_TCP))) {
+		if (!pkt->ip_frag)
+			skb->ip_summed = CHECKSUM_UNNECESSARY;
+		else {
+			__sum16 c = (__force __sum16)pkt->csum;
+			skb->csum = csum_unfold(c);
+			skb->ip_summed = CHECKSUM_COMPLETE;
+		}
+		rxq->stats.rx_cso++;
+	} else
+		skb->ip_summed = CHECKSUM_NONE;
+
+	if (unlikely(pkt->vlan_ex)) {
+		struct vlan_group *grp = pi->vlan_grp;
+
+		rxq->stats.vlan_ex++;
+		if (likely(grp))
+			vlan_hwaccel_receive_skb(skb, grp, ntohs(pkt->vlan));
+		else
+			dev_kfree_skb_any(skb);
+	} else
+		netif_receive_skb(skb);
+
+	return 0;
+
+nomem:	t4_pktgl_free(si);
+	rxq->stats.rx_drops++;
+	return 0;
+}
+
+/**
+ *	restore_rx_bufs - put back a packet's Rx buffers
+ *	@si: the packet gather list
+ *	@q: the SGE free list
+ *	@frags: number of FL buffers to restore
+ *
+ *	Puts back on an FL the Rx buffers associated with @si.  The buffers
+ *	have already been unmapped and are left unmapped, we mark them so to
+ *	prevent further unmapping attempts.
+ *
+ *	This function undoes a series of @unmap_rx_buf calls when we find out
+ *	that the current packet can't be processed right away afterall and we
+ *	need to come back to it later.  This is a very rare event and there's
+ *	no effort to make this particularly efficient.
+ */
+static void restore_rx_bufs(const struct pkt_gl *si, struct sge_fl *q,
+			    int frags)
+{
+	struct rx_sw_desc *d;
+
+	while (frags--) {
+		if (q->cidx == 0)
+			q->cidx = q->size - 1;
+		else
+			q->cidx--;
+		d = &q->sdesc[q->cidx];
+		d->page = si->frags[frags].page;
+		d->dma_addr |= RX_UNMAPPED_BUF;
+		q->avail++;
+	}
+}
+
+/**
+ *	is_new_response - check if a response is newly written
+ *	@r: the response descriptor
+ *	@q: the response queue
+ *
+ *	Returns true if a response descriptor contains a yet unprocessed
+ *	response.
+ */
+static inline bool is_new_response(const struct rsp_ctrl *r,
+				   const struct sge_rspq *q)
+{
+	return RSPD_GEN(r->type_gen) == q->gen;
+}
+
+/**
+ *	rspq_next - advance to the next entry in a response queue
+ *	@q: the queue
+ *
+ *	Updates the state of a response queue to advance it to the next entry.
+ */
+static inline void rspq_next(struct sge_rspq *q)
+{
+	q->cur_desc = (void *)q->cur_desc + q->iqe_len;
+	if (unlikely(++q->cidx == q->size)) {
+		q->cidx = 0;
+		q->gen ^= 1;
+		q->cur_desc = q->desc;
+	}
+}
+
+/**
+ *	process_responses - process responses from an SGE response queue
+ *	@q: the ingress queue to process
+ *	@budget: how many responses can be processed in this round
+ *
+ *	Process responses from an SGE response queue up to the supplied budget.
+ *	Responses include received packets as well as control messages from FW
+ *	or HW.
+ *
+ *	Additionally choose the interrupt holdoff time for the next interrupt
+ *	on this queue.  If the system is under memory shortage use a fairly
+ *	long delay to help recovery.
+ */
+int process_responses(struct sge_rspq *q, int budget)
+{
+	int ret, rsp_type;
+	int budget_left = budget;
+	const struct rsp_ctrl *rc;
+	struct sge_eth_rxq *rxq = container_of(q, struct sge_eth_rxq, rspq);
+
+	while (likely(budget_left)) {
+		rc = (void *)q->cur_desc + (q->iqe_len - sizeof(*rc));
+		if (!is_new_response(rc, q))
+			break;
+
+		rmb();
+		rsp_type = RSPD_TYPE(rc->type_gen);
+		if (likely(rsp_type == RSP_TYPE_FLBUF)) {
+			skb_frag_t *fp;
+			struct pkt_gl si;
+			const struct rx_sw_desc *rsd;
+			u32 len = ntohl(rc->pldbuflen_qid), bufsz, frags;
+
+			if (len & RSPD_NEWBUF) {
+				if (likely(q->offset > 0)) {
+					free_rx_bufs(q->adap, &rxq->fl, 1);
+					q->offset = 0;
+				}
+				len &= RSPD_LEN;
+			}
+			si.tot_len = len;
+
+			/* gather packet fragments */
+			for (frags = 0, fp = si.frags; ; frags++, fp++) {
+				rsd = &rxq->fl.sdesc[rxq->fl.cidx];
+				bufsz = get_buf_size(rsd);
+				fp->page = rsd->page;
+				fp->page_offset = q->offset;
+				fp->size = min(bufsz, len);
+				len -= fp->size;
+				if (!len)
+					break;
+				unmap_rx_buf(q->adap, &rxq->fl);
+			}
+
+			/*
+			 * Last buffer remains mapped so explicitly make it
+			 * coherent for CPU access.
+			 */
+			dma_sync_single_for_cpu(q->adap->pdev_dev,
+						get_buf_addr(rsd),
+						fp->size, DMA_FROM_DEVICE);
+
+			si.va = page_address(si.frags[0].page) +
+				si.frags[0].page_offset;
+			prefetch(si.va);
+
+			si.nfrags = frags + 1;
+			ret = q->handler(q, q->cur_desc, &si);
+			if (likely(ret == 0))
+				q->offset += ALIGN(fp->size, FL_ALIGN);
+			else
+				restore_rx_bufs(&si, &rxq->fl, frags);
+		} else if (likely(rsp_type == RSP_TYPE_CPL)) {
+			ret = q->handler(q, q->cur_desc, NULL);
+		} else {
+			ret = q->handler(q, (const __be64 *)rc, CXGB4_MSG_AN);
+		}
+
+		if (unlikely(ret)) {
+			/* couldn't process descriptor, back off for recovery */
+			q->next_intr_params = QINTR_TIMER_IDX(NOMEM_TMR_IDX);
+			break;
+		}
+
+		rspq_next(q);
+		budget_left--;
+	}
+
+	if (q->offset >= 0 && rxq->fl.size - rxq->fl.avail >= 16)
+		replenish_fl(q->adap, &rxq->fl, GFP_ATOMIC);
+	return budget - budget_left;
+}
+
+/**
+ *	napi_rx_handler - the NAPI handler for Rx processing
+ *	@napi: the napi instance
+ *	@budget: how many packets we can process in this round
+ *
+ *	Handler for new data events when using NAPI.  This does not need any
+ *	locking or protection from interrupts as data interrupts are off at
+ *	this point and other adapter interrupts do not interfere (the latter
+ *	in not a concern at all with MSI-X as non-data interrupts then have
+ *	a separate handler).
+ */
+static int napi_rx_handler(struct napi_struct *napi, int budget)
+{
+	unsigned int params;
+	struct sge_rspq *q = container_of(napi, struct sge_rspq, napi);
+	int work_done = process_responses(q, budget);
+
+	if (likely(work_done < budget)) {
+		napi_complete(napi);
+		params = q->next_intr_params;
+		q->next_intr_params = q->intr_params;
+	} else
+		params = QINTR_TIMER_IDX(7);
+
+	t4_write_reg(q->adap, MYPF_REG(SGE_PF_GTS), CIDXINC(work_done) |
+		     INGRESSQID((u32)q->cntxt_id) | SEINTARM(params));
+	return work_done;
+}
+
+/*
+ * Returns true if a NAPI instance is already scheduled for polling.
+ */
+static inline int napi_is_scheduled(const struct napi_struct *napi)
+{
+	return test_bit(NAPI_STATE_SCHED, &napi->state);
+}
+
+/*
+ * The MSI-X interrupt handler for an SGE response queue.
+ */
+irqreturn_t t4_sge_intr_msix(int irq, void *cookie)
+{
+	struct sge_rspq *q = cookie;
+
+	spin_lock(&q->lock);
+	napi_schedule(&q->napi);
+	spin_unlock(&q->lock);
+	return IRQ_HANDLED;
+}
+
+/*
+ * Process the indirect interrupt entries in the interrupt queue and kick off
+ * NAPI for each queue that has generated an entry.
+ */
+static unsigned int process_intrq(struct adapter *adap)
+{
+	unsigned int credits;
+	const struct rsp_ctrl *rc;
+	struct sge_rspq *q = &adap->sge.intrq;
+
+	for (credits = 0; ; credits++) {
+		rc = (void *)q->cur_desc + (q->iqe_len - sizeof(*rc));
+		if (!is_new_response(rc, q))
+			break;
+
+		rmb();
+		if (RSPD_TYPE(rc->type_gen) == RSP_TYPE_INTR) {
+			unsigned int qid = ntohl(rc->pldbuflen_qid);
+			struct sge_rspq *srcq = adap->sge.ingr_map[qid];
+
+			spin_lock(&srcq->lock);
+			napi_schedule(&srcq->napi);
+			spin_unlock(&srcq->lock);
+		}
+
+		rspq_next(q);
+	}
+
+	t4_write_reg(adap, MYPF_REG(SGE_PF_GTS), CIDXINC(credits) |
+		     INGRESSQID(q->cntxt_id) | SEINTARM(q->intr_params));
+	return credits;
+}
+
+/*
+ * The MSI interrupt handler, which handles data events from SGE response queues
+ * as well as error and other async events as they all use the same MSI vector.
+ */
+static irqreturn_t t4_intr_msi(int irq, void *cookie)
+{
+	struct adapter *adap = cookie;
+
+	t4_slow_intr_handler(adap);
+	process_intrq(adap);
+	return IRQ_HANDLED;
+}
+
+/*
+ * Interrupt handler for legacy INTx interrupts.
+ * Handles data events from SGE response queues as well as error and other
+ * async events as they all use the same interrupt line.
+ */
+static irqreturn_t t4_intr_intx(int irq, void *cookie)
+{
+	struct adapter *adap = cookie;
+
+	t4_write_reg(adap, MYPF_REG(PCIE_PF_CLI), 0);
+	if (t4_slow_intr_handler(adap) | process_intrq(adap))
+		return IRQ_HANDLED;
+	return IRQ_NONE;             /* probably shared interrupt */
+}
+
+/**
+ *	t4_intr_handler - select the top-level interrupt handler
+ *	@adap: the adapter
+ *
+ *	Selects the top-level interrupt handler based on the type of interrupts
+ *	(MSI-X, MSI, or INTx).
+ */
+irq_handler_t t4_intr_handler(struct adapter *adap)
+{
+	if (adap->flags & USING_MSIX)
+		return t4_sge_intr_msix;
+	if (adap->flags & USING_MSI)
+		return t4_intr_msi;
+	return t4_intr_intx;
+}
+
+static void sge_rx_timer_cb(unsigned long data)
+{
+	unsigned long m;
+	unsigned int i, cnt[2];
+	struct adapter *adap = (struct adapter *)data;
+	struct sge *s = &adap->sge;
+
+	for (i = 0; i < ARRAY_SIZE(s->starving_fl); i++)
+		for (m = s->starving_fl[i]; m; m &= m - 1) {
+			struct sge_eth_rxq *rxq;
+			unsigned int id = __ffs(m) + i * BITS_PER_LONG;
+			struct sge_fl *fl = s->egr_map[id];
+
+			rxq = container_of(fl, struct sge_eth_rxq, fl);
+			if (spin_trylock_irq(&rxq->rspq.lock)) {
+				if (!napi_is_scheduled(&rxq->rspq.napi)) {
+					if (fl_starving(fl)) {
+						fl->starving++;
+						__refill_fl(adap, fl);
+					}
+					if (!fl_starving(fl))
+						clear_bit(id, s->starving_fl);
+				}
+				spin_unlock(&rxq->rspq.lock);
+			}
+		}
+
+	t4_write_reg(adap, SGE_DEBUG_INDEX, 13);
+	cnt[0] = t4_read_reg(adap, SGE_DEBUG_DATA_HIGH);
+	cnt[1] = t4_read_reg(adap, SGE_DEBUG_DATA_LOW);
+
+	for (i = 0; i < 2; i++)
+		if (cnt[i] >= s->starve_thres) {
+			if (s->idma_state[i])
+				continue;
+			s->idma_state[i] = 1;
+			t4_write_reg(adap, SGE_DEBUG_INDEX, 11);
+			m = t4_read_reg(adap, SGE_DEBUG_DATA_LOW) >> (i * 16);
+			dev_warn(adap->pdev_dev,
+				 "SGE idma%u starvation detected for "
+				 "queue %lu\n", i, m & 0xffff);
+		} else if (s->idma_state[i])
+			s->idma_state[i] = 0;
+
+	mod_timer(&s->rx_timer, jiffies + RX_QCHECK_PERIOD);
+}
+
+static void sge_tx_timer_cb(unsigned long data)
+{
+	unsigned long m, i;
+	struct adapter *adap = (struct adapter *)data;
+	struct sge *s = &adap->sge;
+
+	for (i = 0; i < ARRAY_SIZE(s->txq_maperr); i++)
+		for (m = s->txq_maperr[i]; m; m &= m - 1) {
+			unsigned long id = __ffs(m) + i * BITS_PER_LONG;
+			struct sge_ofld_txq *txq = s->egr_map[id];
+
+			clear_bit(id, s->txq_maperr);
+			tasklet_schedule(&txq->qresume_tsk);
+		}
+
+	mod_timer(&s->tx_timer, jiffies + TX_QCHECK_PERIOD);
+}
+
+int t4_sge_alloc_rxq(struct adapter *adap, struct sge_rspq *iq, bool fwevtq,
+		     struct net_device *dev, int intr_idx,
+		     struct sge_fl *fl, rspq_handler_t hnd)
+{
+	int ret, flsz = 0;
+	struct fw_iq_cmd c;
+	struct port_info *pi = netdev_priv(dev);
+
+	/* Size needs to be multiple of 16, including status entry. */
+	iq->size = roundup(iq->size, 16);
+
+	iq->desc = alloc_ring(adap->pdev_dev, iq->size, iq->iqe_len, 0,
+			      &iq->phys_addr, NULL, 0);
+	if (!iq->desc)
+		return -ENOMEM;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_IQ_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_WRITE | FW_CMD_EXEC |
+			    FW_IQ_CMD_PFN(0) | FW_IQ_CMD_VFN(0));
+	c.alloc_to_len16 = htonl(FW_IQ_CMD_ALLOC | FW_IQ_CMD_IQSTART(1) |
+				 FW_LEN16(c));
+	c.type_to_iqandstindex = htonl(FW_IQ_CMD_TYPE(FW_IQ_TYPE_FL_INT_CAP) |
+		FW_IQ_CMD_IQASYNCH(fwevtq) | FW_IQ_CMD_VIID(pi->viid) |
+		FW_IQ_CMD_IQANDST(intr_idx < 0) | FW_IQ_CMD_IQANUD(1) |
+		FW_IQ_CMD_IQANDSTINDEX(intr_idx >= 0 ? intr_idx :
+							-intr_idx - 1));
+	c.iqdroprss_to_iqesize = htons(FW_IQ_CMD_IQPCIECH(pi->tx_chan) |
+		FW_IQ_CMD_IQGTSMODE |
+		FW_IQ_CMD_IQINTCNTTHRESH(iq->pktcnt_idx) |
+		FW_IQ_CMD_IQESIZE(ilog2(iq->iqe_len) - 4));
+	c.iqsize = htons(iq->size);
+	c.iqaddr = cpu_to_be64(iq->phys_addr);
+
+	if (fl) {
+		fl->size = roundup(fl->size, 8);
+		fl->desc = alloc_ring(adap->pdev_dev, fl->size, sizeof(__be64),
+				      sizeof(struct rx_sw_desc), &fl->addr,
+				      &fl->sdesc, STAT_LEN);
+		if (!fl->desc)
+			goto fl_nomem;
+
+		flsz = fl->size / 8 + STAT_LEN / sizeof(struct tx_desc);
+		c.iqns_to_fl0congen = htonl(FW_IQ_CMD_FL0PACKEN |
+					    FW_IQ_CMD_FL0PADEN);
+		c.fl0dcaen_to_fl0cidxfthresh = htons(FW_IQ_CMD_FL0FBMIN(2) |
+				FW_IQ_CMD_FL0FBMAX(3));
+		c.fl0size = htons(flsz);
+		c.fl0addr = cpu_to_be64(fl->addr);
+	}
+
+	ret = t4_wr_mbox(adap, 0, &c, sizeof(c), &c);
+	if (ret)
+		goto err;
+
+	spin_lock_init(&iq->lock);
+	netif_napi_add(dev, &iq->napi, napi_rx_handler, 64);
+	iq->cur_desc = iq->desc;
+	iq->cidx = 0;
+	iq->gen = 1;
+	iq->next_intr_params = iq->intr_params;
+	iq->cntxt_id = ntohs(c.iqid);
+	iq->abs_id = ntohs(c.physiqid);
+	iq->size--;                           /* subtract status entry */
+	iq->adap = adap;
+	iq->netdev = dev;
+	iq->handler = hnd;
+
+	/* set offset to -1 to distinguish ingress queues without FL */
+	iq->offset = fl ? 0 : -1;
+
+	adap->sge.ingr_map[iq->cntxt_id] = iq;
+
+	if (fl) {
+		fl->cntxt_id = htons(c.fl0id);
+		fl->avail = fl->pend_cred = 0;
+		fl->pidx = fl->cidx = 0;
+		fl->alloc_failed = fl->large_alloc_failed = fl->starving = 0;
+		adap->sge.egr_map[fl->cntxt_id] = fl;
+		replenish_fl(adap, fl, GFP_KERNEL);
+	}
+	return 0;
+
+fl_nomem:
+	ret = -ENOMEM;
+err:
+	if (iq->desc) {
+		dma_free_coherent(adap->pdev_dev, iq->size * iq->iqe_len,
+				  iq->desc, iq->phys_addr);
+		iq->desc = NULL;
+	}
+	if (fl && fl->desc) {
+		kfree(fl->sdesc);
+		fl->sdesc = NULL;
+		dma_free_coherent(adap->pdev_dev, flsz * sizeof(struct tx_desc),
+				  fl->desc, fl->addr);
+		fl->desc = NULL;
+	}
+	return ret;
+}
+
+static void init_txq(struct adapter *adap, struct sge_txq *q, unsigned int id)
+{
+	q->in_use = 0;
+	q->cidx = q->pidx = 0;
+	q->stops = q->restarts = 0;
+	q->stat = (void *)&q->desc[q->size];
+	q->cntxt_id = id;
+	adap->sge.egr_map[id] = q;
+}
+
+int t4_sge_alloc_eth_txq(struct adapter *adap, struct sge_eth_txq *txq,
+			 struct net_device *dev, struct netdev_queue *netdevq,
+			 unsigned int iqid)
+{
+	int ret, nentries;
+	struct fw_eq_eth_cmd c;
+	struct port_info *pi = netdev_priv(dev);
+
+	/* Add status entries */
+	nentries = txq->q.size + STAT_LEN / sizeof(struct tx_desc);
+
+	txq->q.desc = alloc_ring(adap->pdev_dev, txq->q.size,
+			sizeof(struct tx_desc), sizeof(struct tx_sw_desc),
+			&txq->q.phys_addr, &txq->q.sdesc, STAT_LEN);
+	if (!txq->q.desc)
+		return -ENOMEM;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_EQ_ETH_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_WRITE | FW_CMD_EXEC |
+			    FW_EQ_ETH_CMD_PFN(0) | FW_EQ_ETH_CMD_VFN(0));
+	c.alloc_to_len16 = htonl(FW_EQ_ETH_CMD_ALLOC |
+				 FW_EQ_ETH_CMD_EQSTART | FW_LEN16(c));
+	c.viid_pkd = htonl(FW_EQ_ETH_CMD_VIID(pi->viid));
+	c.fetchszm_to_iqid = htonl(FW_EQ_ETH_CMD_HOSTFCMODE(2) |
+				   FW_EQ_ETH_CMD_PCIECHN(pi->tx_chan) |
+				   FW_EQ_ETH_CMD_IQID(iqid));
+	c.dcaen_to_eqsize = htonl(FW_EQ_ETH_CMD_FBMIN(2) |
+				  FW_EQ_ETH_CMD_FBMAX(3) |
+				  FW_EQ_ETH_CMD_CIDXFTHRESH(5) |
+				  FW_EQ_ETH_CMD_EQSIZE(nentries));
+	c.eqaddr = cpu_to_be64(txq->q.phys_addr);
+
+	ret = t4_wr_mbox(adap, 0, &c, sizeof(c), &c);
+	if (ret) {
+		kfree(txq->q.sdesc);
+		txq->q.sdesc = NULL;
+		dma_free_coherent(adap->pdev_dev,
+				  nentries * sizeof(struct tx_desc),
+				  txq->q.desc, txq->q.phys_addr);
+		txq->q.desc = NULL;
+		return ret;
+	}
+
+	init_txq(adap, &txq->q, FW_EQ_ETH_CMD_EQID_GET(ntohl(c.eqid_pkd)));
+	txq->txq = netdevq;
+	txq->tso = txq->tx_cso = txq->vlan_ins = 0;
+	txq->mapping_err = 0;
+	return 0;
+}
+
+int t4_sge_alloc_ctrl_txq(struct adapter *adap, struct sge_ctrl_txq *txq,
+			  struct net_device *dev, unsigned int iqid,
+			  unsigned int cmplqid)
+{
+	int ret, nentries;
+	struct fw_eq_ctrl_cmd c;
+	struct port_info *pi = netdev_priv(dev);
+
+	/* Add status entries */
+	nentries = txq->q.size + STAT_LEN / sizeof(struct tx_desc);
+
+	txq->q.desc = alloc_ring(adap->pdev_dev, nentries,
+				 sizeof(struct tx_desc), 0, &txq->q.phys_addr,
+				 NULL, 0);
+	if (!txq->q.desc)
+		return -ENOMEM;
+
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_EQ_CTRL_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_WRITE | FW_CMD_EXEC |
+			    FW_EQ_CTRL_CMD_PFN(0) | FW_EQ_CTRL_CMD_VFN(0));
+	c.alloc_to_len16 = htonl(FW_EQ_CTRL_CMD_ALLOC |
+				 FW_EQ_CTRL_CMD_EQSTART | FW_LEN16(c));
+	c.cmpliqid_eqid = htonl(FW_EQ_CTRL_CMD_CMPLIQID(cmplqid));
+	c.physeqid_pkd = htonl(0);
+	c.fetchszm_to_iqid = htonl(FW_EQ_CTRL_CMD_HOSTFCMODE(2) |
+				   FW_EQ_CTRL_CMD_PCIECHN(pi->tx_chan) |
+				   FW_EQ_CTRL_CMD_IQID(iqid));
+	c.dcaen_to_eqsize = htonl(FW_EQ_CTRL_CMD_FBMIN(2) |
+				  FW_EQ_CTRL_CMD_FBMAX(3) |
+				  FW_EQ_CTRL_CMD_CIDXFTHRESH(5) |
+				  FW_EQ_CTRL_CMD_EQSIZE(nentries));
+	c.eqaddr = cpu_to_be64(txq->q.phys_addr);
+
+	ret = t4_wr_mbox(adap, 0, &c, sizeof(c), &c);
+	if (ret) {
+		dma_free_coherent(adap->pdev_dev,
+				  nentries * sizeof(struct tx_desc),
+				  txq->q.desc, txq->q.phys_addr);
+		txq->q.desc = NULL;
+		return ret;
+	}
+
+	init_txq(adap, &txq->q, FW_EQ_CTRL_CMD_EQID_GET(ntohl(c.cmpliqid_eqid)));
+	txq->adap = adap;
+	skb_queue_head_init(&txq->sendq);
+	tasklet_init(&txq->qresume_tsk, restart_ctrlq, (unsigned long)txq);
+	txq->full = 0;
+	return 0;
+}
+
+int t4_sge_alloc_ofld_txq(struct adapter *adap, struct sge_ofld_txq *txq,
+			  struct net_device *dev, unsigned int iqid)
+{
+	int ret, nentries;
+	struct fw_eq_ofld_cmd c;
+	struct port_info *pi = netdev_priv(dev);
+
+	/* Add status entries */
+	nentries = txq->q.size + STAT_LEN / sizeof(struct tx_desc);
+
+	txq->q.desc = alloc_ring(adap->pdev_dev, txq->q.size,
+			sizeof(struct tx_desc), sizeof(struct tx_sw_desc),
+			&txq->q.phys_addr, &txq->q.sdesc, STAT_LEN);
+	if (!txq->q.desc)
+		return -ENOMEM;
+
+	memset(&c, 0, sizeof(c));
+	c.op_to_vfn = htonl(FW_CMD_OP(FW_EQ_OFLD_CMD) | FW_CMD_REQUEST |
+			    FW_CMD_WRITE | FW_CMD_EXEC |
+			    FW_EQ_OFLD_CMD_PFN(0) | FW_EQ_OFLD_CMD_VFN(0));
+	c.alloc_to_len16 = htonl(FW_EQ_OFLD_CMD_ALLOC |
+				 FW_EQ_OFLD_CMD_EQSTART | FW_LEN16(c));
+	c.fetchszm_to_iqid = htonl(FW_EQ_OFLD_CMD_HOSTFCMODE(2) |
+				   FW_EQ_OFLD_CMD_PCIECHN(pi->tx_chan) |
+				   FW_EQ_OFLD_CMD_IQID(iqid));
+	c.dcaen_to_eqsize = htonl(FW_EQ_OFLD_CMD_FBMIN(2) |
+				  FW_EQ_OFLD_CMD_FBMAX(3) |
+				  FW_EQ_OFLD_CMD_CIDXFTHRESH(5) |
+				  FW_EQ_OFLD_CMD_EQSIZE(nentries));
+	c.eqaddr = cpu_to_be64(txq->q.phys_addr);
+
+	ret = t4_wr_mbox(adap, 0, &c, sizeof(c), &c);
+	if (ret) {
+		kfree(txq->q.sdesc);
+		txq->q.sdesc = NULL;
+		dma_free_coherent(adap->pdev_dev,
+				  nentries * sizeof(struct tx_desc),
+				  txq->q.desc, txq->q.phys_addr);
+		txq->q.desc = NULL;
+		return ret;
+	}
+
+	init_txq(adap, &txq->q, FW_EQ_OFLD_CMD_EQID_GET(ntohl(c.eqid_pkd)));
+	txq->adap = adap;
+	skb_queue_head_init(&txq->sendq);
+	tasklet_init(&txq->qresume_tsk, restart_ofldq, (unsigned long)txq);
+	txq->full = 0;
+	txq->mapping_err = 0;
+	return 0;
+}
+
+static void free_txq(struct adapter *adap, struct sge_txq *q)
+{
+	dma_free_coherent(adap->pdev_dev,
+			  q->size * sizeof(struct tx_desc) + STAT_LEN,
+			  q->desc, q->phys_addr);
+	q->cntxt_id = 0;
+	q->sdesc = NULL;
+	q->desc = NULL;
+}
+
+static void free_rspq_fl(struct adapter *adap, struct sge_rspq *rq,
+			 struct sge_fl *fl)
+{
+	unsigned int fl_id = fl ? fl->cntxt_id : 0xffff;
+
+	t4_iq_free(adap, 0, 0, 0, FW_IQ_TYPE_FL_INT_CAP, rq->cntxt_id, fl_id,
+		   0xffff);
+	dma_free_coherent(adap->pdev_dev, (rq->size + 1) * rq->iqe_len,
+			  rq->desc, rq->phys_addr);
+	netif_napi_del(&rq->napi);
+	rq->netdev = NULL;
+	rq->cntxt_id = rq->abs_id = 0;
+	rq->desc = NULL;
+
+	if (fl) {
+		free_rx_bufs(adap, fl, fl->avail);
+		dma_free_coherent(adap->pdev_dev, fl->size * 8 + STAT_LEN,
+				  fl->desc, fl->addr);
+		kfree(fl->sdesc);
+		fl->sdesc = NULL;
+		fl->cntxt_id = 0;
+		fl->desc = NULL;
+	}
+}
+
+/**
+ *	t4_free_sge_resources - free SGE resources
+ *	@adap: the adapter
+ *
+ *	Frees resources used by the SGE queue sets.
+ */
+void t4_free_sge_resources(struct adapter *adap)
+{
+	int i;
+	struct sge_eth_rxq *eq = adap->sge.ethrxq;
+	struct sge_eth_txq *etq = adap->sge.ethtxq;
+	struct sge_ofld_rxq *oq = adap->sge.ofldrxq;
+	struct sge_rspq *intrq = &adap->sge.intrq;
+
+	/* clean up Ethernet Tx/Rx queues */
+	for (i = 0; i < adap->sge.ethqsets; i++, eq++, etq++) {
+		if (eq->rspq.desc)
+			free_rspq_fl(adap, &eq->rspq, &eq->fl);
+		if (etq->q.desc) {
+			t4_eth_eq_free(adap, 0, 0, 0, etq->q.cntxt_id);
+			free_tx_desc(adap, &etq->q, etq->q.in_use, true);
+			kfree(etq->q.sdesc);
+			free_txq(adap, &etq->q);
+		}
+	}
+
+	/* clean up RDMA and iSCSI Rx queues */
+	for (i = 0; i < adap->sge.ofldqsets; i++, oq++) {
+		if (oq->rspq.desc)
+			free_rspq_fl(adap, &oq->rspq, &oq->fl);
+	}
+	for (i = 0, oq = adap->sge.rdmarxq; i < adap->sge.rdmaqs; i++, oq++) {
+		if (oq->rspq.desc)
+			free_rspq_fl(adap, &oq->rspq, &oq->fl);
+	}
+
+	/* clean up offload Tx queues */
+	for (i = 0; i < ARRAY_SIZE(adap->sge.ofldtxq); i++) {
+		struct sge_ofld_txq *q = &adap->sge.ofldtxq[i];
+
+		if (q->q.desc) {
+			tasklet_kill(&q->qresume_tsk);
+			t4_ofld_eq_free(adap, 0, 0, 0, q->q.cntxt_id);
+			free_tx_desc(adap, &q->q, q->q.in_use, false);
+			kfree(q->q.sdesc);
+			__skb_queue_purge(&q->sendq);
+			free_txq(adap, &q->q);
+		}
+	}
+
+	/* clean up control Tx queues */
+	for (i = 0; i < ARRAY_SIZE(adap->sge.ctrlq); i++) {
+		struct sge_ctrl_txq *cq = &adap->sge.ctrlq[i];
+
+		if (cq->q.desc) {
+			tasklet_kill(&cq->qresume_tsk);
+			t4_ctrl_eq_free(adap, 0, 0, 0, cq->q.cntxt_id);
+			__skb_queue_purge(&cq->sendq);
+			free_txq(adap, &cq->q);
+		}
+	}
+
+	if (adap->sge.fw_evtq.desc)
+		free_rspq_fl(adap, &adap->sge.fw_evtq, NULL);
+
+	if (intrq->desc) {
+		t4_iq_free(adap, 0, 0, 0, FW_IQ_TYPE_FL_INT_CAP,
+			   intrq->cntxt_id, 0xffff, 0xffff);
+		dma_free_coherent(adap->pdev_dev,
+				  (intrq->size + 1) * intrq->iqe_len,
+				  intrq->desc, intrq->phys_addr);
+		intrq->cntxt_id = intrq->abs_id = 0;
+		intrq->desc = NULL;
+		/* this queue doesn't use NAPI */
+	}
+}
+
+void t4_sge_start(struct adapter *adap)
+{
+	mod_timer(&adap->sge.rx_timer, jiffies + RX_QCHECK_PERIOD);
+	mod_timer(&adap->sge.tx_timer, jiffies + TX_QCHECK_PERIOD);
+}
+
+/**
+ *	t4_sge_stop - disable SGE operation
+ *	@adap: the adapter
+ *
+ *	Stop tasklets and timers associated with the DMA engine.  Note that
+ *	this is effective only if measures have been taken to disable any HW
+ *	events that may restart them.
+ */
+void t4_sge_stop(struct adapter *adap)
+{
+	int i;
+	struct sge *s = &adap->sge;
+
+	if (in_interrupt())  /* actions below require waiting */
+		return;
+
+	if (s->rx_timer.function)
+		del_timer_sync(&s->rx_timer);
+	if (s->tx_timer.function)
+		del_timer_sync(&s->tx_timer);
+
+	for (i = 0; i < ARRAY_SIZE(s->ofldtxq); i++) {
+		struct sge_ofld_txq *q = &s->ofldtxq[i];
+
+		if (q->q.desc)
+			tasklet_kill(&q->qresume_tsk);
+	}
+	for (i = 0; i < ARRAY_SIZE(s->ctrlq); i++) {
+		struct sge_ctrl_txq *cq = &s->ctrlq[i];
+
+		if (cq->q.desc)
+			tasklet_kill(&cq->qresume_tsk);
+	}
+}
+
+/**
+ *	t4_sge_init - initialize SGE
+ *	@adap: the adapter
+ *
+ *	Performs SGE initialization needed every time after a chip reset.
+ *	We do not initialize any of the queues here, instead the driver
+ *	top-level must request them individually.
+ */
+void t4_sge_init(struct adapter *adap)
+{
+	struct sge *s = &adap->sge;
+	unsigned int fl_align_log = ilog2(FL_ALIGN);
+
+	t4_set_reg_field(adap, SGE_CONTROL, PKTSHIFT_MASK |
+			 INGPADBOUNDARY_MASK | EGRSTATUSPAGESIZE,
+			 INGPADBOUNDARY(fl_align_log - 5) | PKTSHIFT(2) |
+			 RXPKTCPLMODE |
+			 (STAT_LEN == 128 ? EGRSTATUSPAGESIZE : 0));
+	t4_set_reg_field(adap, SGE_HOST_PAGE_SIZE, HOSTPAGESIZEPF0_MASK,
+			 HOSTPAGESIZEPF0(PAGE_SHIFT - 10));
+	t4_write_reg(adap, SGE_FL_BUFFER_SIZE0, PAGE_SIZE);
+#if FL_PG_ORDER > 0
+	t4_write_reg(adap, SGE_FL_BUFFER_SIZE1, PAGE_SIZE << FL_PG_ORDER);
+#endif
+	t4_write_reg(adap, SGE_INGRESS_RX_THRESHOLD,
+		     THRESHOLD_0(s->counter_val[0]) |
+		     THRESHOLD_1(s->counter_val[1]) |
+		     THRESHOLD_2(s->counter_val[2]) |
+		     THRESHOLD_3(s->counter_val[3]));
+	t4_write_reg(adap, SGE_TIMER_VALUE_0_AND_1,
+		     TIMERVALUE0(us_to_core_ticks(adap, s->timer_val[0])) |
+		     TIMERVALUE1(us_to_core_ticks(adap, s->timer_val[1])));
+	t4_write_reg(adap, SGE_TIMER_VALUE_2_AND_3,
+		     TIMERVALUE0(us_to_core_ticks(adap, s->timer_val[2])) |
+		     TIMERVALUE1(us_to_core_ticks(adap, s->timer_val[3])));
+	t4_write_reg(adap, SGE_TIMER_VALUE_4_AND_5,
+		     TIMERVALUE0(us_to_core_ticks(adap, s->timer_val[4])) |
+		     TIMERVALUE1(us_to_core_ticks(adap, s->timer_val[5])));
+	setup_timer(&s->rx_timer, sge_rx_timer_cb, (unsigned long)adap);
+	setup_timer(&s->tx_timer, sge_tx_timer_cb, (unsigned long)adap);
+	s->starve_thres = core_ticks_per_usec(adap) * 1000000;  /* 1 s */
+	s->idma_state[0] = s->idma_state[1] = 0;
+}
-- 
1.5.4