bnx2x_cmn.c

/* bnx2x_cmn.c: QLogic Everest network driver.
 *
 * Copyright (c) 2007-2013 Broadcom Corporation
 * Copyright (c) 2014 QLogic Corporation
 * All rights reserved
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation.
 *
 * Maintained by: Ariel Elior <ariel.elior@qlogic.com>
 * Written by: Eliezer Tamir
 * Based on code from Michael Chan's bnx2 driver
 * UDP CSUM errata workaround by Arik Gendelman
 * Slowpath and fastpath rework by Vladislav Zolotarov
 * Statistics and Link management by Yitchak Gertner
 *
 */

#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt

#include <linux/etherdevice.h>
#include <linux/if_vlan.h>
#include <linux/interrupt.h>
#include <linux/ip.h>
#include <linux/crash_dump.h>
#include <net/tcp.h>
#include <net/ipv6.h>
#include <net/ip6_checksum.h>
#include <net/busy_poll.h>
#include <linux/prefetch.h>
#include "bnx2x_cmn.h"
#include "bnx2x_init.h"
#include "bnx2x_sp.h"

#ifdef CONFIG_XEN_SME
#include <linux/irqdesc.h>
extern int (*lnk_intercept_sk_buff) (struct sk_buff *skb, char *extra_dbg_string);
extern int (*lnk_intercept_rx_path) (void *dev, int fp_idx,
		uint16_t rx_bd_prod, uint16_t rx_bd_cons,
		uint16_t comp_prod, uint16_t comp_cons, char *extra_dbg_string);
#define SME_NAPI_POLL_WEIGHT 64
#endif

static void bnx2x_free_fp_mem_cnic(struct bnx2x *bp);
static int bnx2x_alloc_fp_mem_cnic(struct bnx2x *bp);
static int bnx2x_alloc_fp_mem(struct bnx2x *bp);
static int bnx2x_poll(struct napi_struct *napi, int budget);

static void bnx2x_add_all_napi_cnic(struct bnx2x *bp)
{
	int i;

	/* Add NAPI objects */
	for_each_rx_queue_cnic(bp, i) {
		netif_napi_add(bp->dev, &bnx2x_fp(bp, i, napi),
			       bnx2x_poll, NAPI_POLL_WEIGHT);
	}
}

static void bnx2x_add_all_napi(struct bnx2x *bp)
{
	int i;

	/* Add NAPI objects */
	for_each_eth_queue(bp, i) {
#if CONFIG_XEN_SME
		netif_napi_add(bp->dev, &bnx2x_fp(bp, i, napi),
				bnx2x_poll, SME_NAPI_POLL_WEIGHT);
#else
		netif_napi_add(bp->dev, &bnx2x_fp(bp, i, napi),
			       bnx2x_poll, NAPI_POLL_WEIGHT);
#endif
	}
}

static int bnx2x_calc_num_queues(struct bnx2x *bp)
{
	int nq = bnx2x_num_queues ? : netif_get_num_default_rss_queues();

	/* Reduce memory usage in kdump environment by using only one queue */
	if (is_kdump_kernel())
		nq = 1;

	nq = clamp(nq, 1, BNX2X_MAX_QUEUES(bp));
	return nq;
}

/**
 * bnx2x_move_fp - move content of the fastpath structure.
 *
 * @bp:		driver handle
 * @from:	source FP index
 * @to:		destination FP index
 *
 * Makes sure the contents of the bp->fp[to].napi is kept
 * intact. This is done by first copying the napi struct from
 * the target to the source, and then mem copying the entire
 * source onto the target. Update txdata pointers and related
 * content.
 */
static inline void bnx2x_move_fp(struct bnx2x *bp, int from, int to)
{
	struct bnx2x_fastpath *from_fp = &bp->fp[from];
	struct bnx2x_fastpath *to_fp = &bp->fp[to];
	struct bnx2x_sp_objs *from_sp_objs = &bp->sp_objs[from];
	struct bnx2x_sp_objs *to_sp_objs = &bp->sp_objs[to];
	struct bnx2x_fp_stats *from_fp_stats = &bp->fp_stats[from];
	struct bnx2x_fp_stats *to_fp_stats = &bp->fp_stats[to];
	int old_max_eth_txqs, new_max_eth_txqs;
	int old_txdata_index = 0, new_txdata_index = 0;
	struct bnx2x_agg_info *old_tpa_info = to_fp->tpa_info;

	/* Copy the NAPI object as it has been already initialized */
	from_fp->napi = to_fp->napi;

	/* Move bnx2x_fastpath contents */
	memcpy(to_fp, from_fp, sizeof(*to_fp));
	to_fp->index = to;

	/* Retain the tpa_info of the original `to' version as we don't want
	 * 2 FPs to contain the same tpa_info pointer.
	 */
	to_fp->tpa_info = old_tpa_info;

	/* move sp_objs contents as well, as their indices match fp ones */
	memcpy(to_sp_objs, from_sp_objs, sizeof(*to_sp_objs));

	/* move fp_stats contents as well, as their indices match fp ones */
	memcpy(to_fp_stats, from_fp_stats, sizeof(*to_fp_stats));

	/* Update txdata pointers in fp and move txdata content accordingly:
	 * Each fp consumes 'max_cos' txdata structures, so the index should be
	 * decremented by max_cos x delta.
	 */

	old_max_eth_txqs = BNX2X_NUM_ETH_QUEUES(bp) * (bp)->max_cos;
	new_max_eth_txqs = (BNX2X_NUM_ETH_QUEUES(bp) - from + to) *
				(bp)->max_cos;
	if (from == FCOE_IDX(bp)) {
		old_txdata_index = old_max_eth_txqs + FCOE_TXQ_IDX_OFFSET;
		new_txdata_index = new_max_eth_txqs + FCOE_TXQ_IDX_OFFSET;
	}

	memcpy(&bp->bnx2x_txq[new_txdata_index],
	       &bp->bnx2x_txq[old_txdata_index],
	       sizeof(struct bnx2x_fp_txdata));
	to_fp->txdata_ptr[0] = &bp->bnx2x_txq[new_txdata_index];
}

/**
 * bnx2x_fill_fw_str - Fill buffer with FW version string.
 *
 * @bp:        driver handle
 * @buf:       character buffer to fill with the fw name
 * @buf_len:   length of the above buffer
 *
 */
void bnx2x_fill_fw_str(struct bnx2x *bp, char *buf, size_t buf_len)
{
	if (IS_PF(bp)) {
		u8 phy_fw_ver[PHY_FW_VER_LEN];

		phy_fw_ver[0] = '\0';
		bnx2x_get_ext_phy_fw_version(&bp->link_params,
					     phy_fw_ver, PHY_FW_VER_LEN);
		strlcpy(buf, bp->fw_ver, buf_len);
		snprintf(buf + strlen(bp->fw_ver), 32 - strlen(bp->fw_ver),
			 "bc %d.%d.%d%s%s",
			 (bp->common.bc_ver & 0xff0000) >> 16,
			 (bp->common.bc_ver & 0xff00) >> 8,
			 (bp->common.bc_ver & 0xff),
			 ((phy_fw_ver[0] != '\0') ? " phy " : ""), phy_fw_ver);
	} else {
		bnx2x_vf_fill_fw_str(bp, buf, buf_len);
	}
}

/**
 * bnx2x_shrink_eth_fp - guarantees fastpath structures stay intact
 *
 * @bp:	driver handle
 * @delta:	number of eth queues which were not allocated
 */
static void bnx2x_shrink_eth_fp(struct bnx2x *bp, int delta)
{
	int i, cos, old_eth_num = BNX2X_NUM_ETH_QUEUES(bp);

	/* Queue pointer cannot be re-set on an fp-basis, as moving pointer
	 * backward along the array could cause memory to be overridden
	 */
	for (cos = 1; cos < bp->max_cos; cos++) {
		for (i = 0; i < old_eth_num - delta; i++) {
			struct bnx2x_fastpath *fp = &bp->fp[i];
			int new_idx = cos * (old_eth_num - delta) + i;

			memcpy(&bp->bnx2x_txq[new_idx], fp->txdata_ptr[cos],
			       sizeof(struct bnx2x_fp_txdata));
			fp->txdata_ptr[cos] = &bp->bnx2x_txq[new_idx];
		}
	}
}

int bnx2x_load_count[2][3] = { {0} }; /* per-path: 0-common, 1-port0, 2-port1 */

/* free skb in the packet ring at pos idx
 * return idx of last bd freed
 */
static u16 bnx2x_free_tx_pkt(struct bnx2x *bp, struct bnx2x_fp_txdata *txdata,
			     u16 idx, unsigned int *pkts_compl,
			     unsigned int *bytes_compl)
{
	struct sw_tx_bd *tx_buf = &txdata->tx_buf_ring[idx];
	struct eth_tx_start_bd *tx_start_bd;
	struct eth_tx_bd *tx_data_bd;
	struct sk_buff *skb = tx_buf->skb;
	u16 bd_idx = TX_BD(tx_buf->first_bd), new_cons;
	int nbd;
	u16 split_bd_len = 0;

	/* prefetch skb end pointer to speedup dev_kfree_skb() */
	prefetch(&skb->end);

	DP(NETIF_MSG_TX_DONE, "fp[%d]: pkt_idx %d  buff @(%p)->skb %p\n",
	   txdata->txq_index, idx, tx_buf, skb);

	tx_start_bd = &txdata->tx_desc_ring[bd_idx].start_bd;

	nbd = le16_to_cpu(tx_start_bd->nbd) - 1;
#ifdef BNX2X_STOP_ON_ERROR
	if ((nbd - 1) > (MAX_SKB_FRAGS + 2)) {
		BNX2X_ERR("BAD nbd!\n");
		bnx2x_panic();
	}
#endif
	new_cons = nbd + tx_buf->first_bd;

	/* Get the next bd */
	bd_idx = TX_BD(NEXT_TX_IDX(bd_idx));

	/* Skip a parse bd... */
	--nbd;
	bd_idx = TX_BD(NEXT_TX_IDX(bd_idx));

	if (tx_buf->flags & BNX2X_HAS_SECOND_PBD) {
		/* Skip second parse bd... */
		--nbd;
		bd_idx = TX_BD(NEXT_TX_IDX(bd_idx));
	}

	/* TSO headers+data bds share a common mapping. See bnx2x_tx_split() */
	if (tx_buf->flags & BNX2X_TSO_SPLIT_BD) {
		tx_data_bd = &txdata->tx_desc_ring[bd_idx].reg_bd;
		split_bd_len = BD_UNMAP_LEN(tx_data_bd);
		--nbd;
		bd_idx = TX_BD(NEXT_TX_IDX(bd_idx));
	}

	/* unmap first bd */
	dma_unmap_single(&bp->pdev->dev, BD_UNMAP_ADDR(tx_start_bd),
			 BD_UNMAP_LEN(tx_start_bd) + split_bd_len,
			 DMA_TO_DEVICE);

	/* now free frags */
	while (nbd > 0) {

		tx_data_bd = &txdata->tx_desc_ring[bd_idx].reg_bd;
		dma_unmap_page(&bp->pdev->dev, BD_UNMAP_ADDR(tx_data_bd),
			       BD_UNMAP_LEN(tx_data_bd), DMA_TO_DEVICE);
		if (--nbd)
			bd_idx = TX_BD(NEXT_TX_IDX(bd_idx));
	}

	/* release skb */
	WARN_ON(!skb);
	if (likely(skb)) {
		(*pkts_compl)++;
		(*bytes_compl) += skb->len;
		dev_kfree_skb_any(skb);
	}

	tx_buf->first_bd = 0;
	tx_buf->skb = NULL;

	return new_cons;
}

int bnx2x_tx_int(struct bnx2x *bp, struct bnx2x_fp_txdata *txdata)
{
	struct netdev_queue *txq;
	u16 hw_cons, sw_cons, bd_cons = txdata->tx_bd_cons;
	unsigned int pkts_compl = 0, bytes_compl = 0;

#ifdef BNX2X_STOP_ON_ERROR
	if (unlikely(bp->panic))
		return -1;
#endif

	txq = netdev_get_tx_queue(bp->dev, txdata->txq_index);
	hw_cons = le16_to_cpu(*txdata->tx_cons_sb);
	sw_cons = txdata->tx_pkt_cons;

	while (sw_cons != hw_cons) {
		u16 pkt_cons;

		pkt_cons = TX_BD(sw_cons);

		DP(NETIF_MSG_TX_DONE,
		   "queue[%d]: hw_cons %u  sw_cons %u  pkt_cons %u\n",
		   txdata->txq_index, hw_cons, sw_cons, pkt_cons);

		bd_cons = bnx2x_free_tx_pkt(bp, txdata, pkt_cons,
					    &pkts_compl, &bytes_compl);

		sw_cons++;
	}

	netdev_tx_completed_queue(txq, pkts_compl, bytes_compl);

	txdata->tx_pkt_cons = sw_cons;
	txdata->tx_bd_cons = bd_cons;

	/* Need to make the tx_bd_cons update visible to start_xmit()
	 * before checking for netif_tx_queue_stopped().  Without the
	 * memory barrier, there is a small possibility that
	 * start_xmit() will miss it and cause the queue to be stopped
	 * forever.
	 * On the other hand we need an rmb() here to ensure the proper
	 * ordering of bit testing in the following
	 * netif_tx_queue_stopped(txq) call.
	 */
	smp_mb();

	if (unlikely(netif_tx_queue_stopped(txq))) {
		/* Taking tx_lock() is needed to prevent re-enabling the queue
		 * while it's empty. This could have happen if rx_action() gets
		 * suspended in bnx2x_tx_int() after the condition before
		 * netif_tx_wake_queue(), while tx_action (bnx2x_start_xmit()):
		 *
		 * stops the queue->sees fresh tx_bd_cons->releases the queue->
		 * sends some packets consuming the whole queue again->
		 * stops the queue
		 */

		__netif_tx_lock(txq, smp_processor_id());

		if ((netif_tx_queue_stopped(txq)) &&
		    (bp->state == BNX2X_STATE_OPEN) &&
		    (bnx2x_tx_avail(bp, txdata) >= MAX_DESC_PER_TX_PKT))
			netif_tx_wake_queue(txq);

		__netif_tx_unlock(txq);
	}
	return 0;
}

static inline void bnx2x_update_last_max_sge(struct bnx2x_fastpath *fp,
					     u16 idx)
{
	u16 last_max = fp->last_max_sge;

	if (SUB_S16(idx, last_max) > 0)
		fp->last_max_sge = idx;
}

static inline void bnx2x_update_sge_prod(struct bnx2x_fastpath *fp,
					 u16 sge_len,
					 struct eth_end_agg_rx_cqe *cqe)
{
	struct bnx2x *bp = fp->bp;
	u16 last_max, last_elem, first_elem;
	u16 delta = 0;
	u16 i;

	if (!sge_len)
		return;

	/* First mark all used pages */
	for (i = 0; i < sge_len; i++)
		BIT_VEC64_CLEAR_BIT(fp->sge_mask,
			RX_SGE(le16_to_cpu(cqe->sgl_or_raw_data.sgl[i])));

	DP(NETIF_MSG_RX_STATUS, "fp_cqe->sgl[%d] = %d\n",
	   sge_len - 1, le16_to_cpu(cqe->sgl_or_raw_data.sgl[sge_len - 1]));

	/* Here we assume that the last SGE index is the biggest */
	prefetch((void *)(fp->sge_mask));
	bnx2x_update_last_max_sge(fp,
		le16_to_cpu(cqe->sgl_or_raw_data.sgl[sge_len - 1]));

	last_max = RX_SGE(fp->last_max_sge);
	last_elem = last_max >> BIT_VEC64_ELEM_SHIFT;
	first_elem = RX_SGE(fp->rx_sge_prod) >> BIT_VEC64_ELEM_SHIFT;

	/* If ring is not full */
	if (last_elem + 1 != first_elem)
		last_elem++;

	/* Now update the prod */
	for (i = first_elem; i != last_elem; i = NEXT_SGE_MASK_ELEM(i)) {
		if (likely(fp->sge_mask[i]))
			break;

		fp->sge_mask[i] = BIT_VEC64_ELEM_ONE_MASK;
		delta += BIT_VEC64_ELEM_SZ;
	}

	if (delta > 0) {
		fp->rx_sge_prod += delta;
		/* clear page-end entries */
		bnx2x_clear_sge_mask_next_elems(fp);
	}

	DP(NETIF_MSG_RX_STATUS,
	   "fp->last_max_sge = %d  fp->rx_sge_prod = %d\n",
	   fp->last_max_sge, fp->rx_sge_prod);
}

/* Get Toeplitz hash value in the skb using the value from the
 * CQE (calculated by HW).
 */
static u32 bnx2x_get_rxhash(const struct bnx2x *bp,
			    const struct eth_fast_path_rx_cqe *cqe,
			    enum pkt_hash_types *rxhash_type)
{
	/* Get Toeplitz hash from CQE */
	if ((bp->dev->features & NETIF_F_RXHASH) &&
	    (cqe->status_flags & ETH_FAST_PATH_RX_CQE_RSS_HASH_FLG)) {
		enum eth_rss_hash_type htype;

		htype = cqe->status_flags & ETH_FAST_PATH_RX_CQE_RSS_HASH_TYPE;
		*rxhash_type = ((htype == TCP_IPV4_HASH_TYPE) ||
				(htype == TCP_IPV6_HASH_TYPE)) ?
			       PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3;

		return le32_to_cpu(cqe->rss_hash_result);
	}
	*rxhash_type = PKT_HASH_TYPE_NONE;
	return 0;
}

static void bnx2x_tpa_start(struct bnx2x_fastpath *fp, u16 queue,
			    u16 cons, u16 prod,
			    struct eth_fast_path_rx_cqe *cqe)
{
	struct bnx2x *bp = fp->bp;
	struct sw_rx_bd *cons_rx_buf = &fp->rx_buf_ring[cons];
	struct sw_rx_bd *prod_rx_buf = &fp->rx_buf_ring[prod];
	struct eth_rx_bd *prod_bd = &fp->rx_desc_ring[prod];
	dma_addr_t mapping;
	struct bnx2x_agg_info *tpa_info = &fp->tpa_info[queue];
	struct sw_rx_bd *first_buf = &tpa_info->first_buf;

	/* print error if current state != stop */
	if (tpa_info->tpa_state != BNX2X_TPA_STOP)
		BNX2X_ERR("start of bin not in stop [%d]\n", queue);

	/* Try to map an empty data buffer from the aggregation info  */
	mapping = dma_map_single(&bp->pdev->dev,
				 first_buf->data + NET_SKB_PAD,
				 fp->rx_buf_size, DMA_FROM_DEVICE);
	/*
	 *  ...if it fails - move the skb from the consumer to the producer
	 *  and set the current aggregation state as ERROR to drop it
	 *  when TPA_STOP arrives.
	 */

	if (unlikely(dma_mapping_error(&bp->pdev->dev, mapping))) {
		/* Move the BD from the consumer to the producer */
		bnx2x_reuse_rx_data(fp, cons, prod);
		tpa_info->tpa_state = BNX2X_TPA_ERROR;
		return;
	}

	/* move empty data from pool to prod */
	prod_rx_buf->data = first_buf->data;
	dma_unmap_addr_set(prod_rx_buf, mapping, mapping);
	/* point prod_bd to new data */
	prod_bd->addr_hi = cpu_to_le32(U64_HI(mapping));
	prod_bd->addr_lo = cpu_to_le32(U64_LO(mapping));

	/* move partial skb from cons to pool (don't unmap yet) */
	*first_buf = *cons_rx_buf;

	/* mark bin state as START */
	tpa_info->parsing_flags =
		le16_to_cpu(cqe->pars_flags.flags);
	tpa_info->vlan_tag = le16_to_cpu(cqe->vlan_tag);
	tpa_info->tpa_state = BNX2X_TPA_START;
	tpa_info->len_on_bd = le16_to_cpu(cqe->len_on_bd);
	tpa_info->placement_offset = cqe->placement_offset;
	tpa_info->rxhash = bnx2x_get_rxhash(bp, cqe, &tpa_info->rxhash_type);
	if (fp->mode == TPA_MODE_GRO) {
		u16 gro_size = le16_to_cpu(cqe->pkt_len_or_gro_seg_len);
		tpa_info->full_page = SGE_PAGES / gro_size * gro_size;
		tpa_info->gro_size = gro_size;
	}

#ifdef BNX2X_STOP_ON_ERROR
	fp->tpa_queue_used |= (1 << queue);
	DP(NETIF_MSG_RX_STATUS, "fp->tpa_queue_used = 0x%llx\n",
	   fp->tpa_queue_used);
#endif
}

/* Timestamp option length allowed for TPA aggregation:
 *
 *		nop nop kind length echo val
 */
#define TPA_TSTAMP_OPT_LEN	12
/**
 * bnx2x_set_gro_params - compute GRO values
 *
 * @skb:		packet skb
 * @parsing_flags:	parsing flags from the START CQE
 * @len_on_bd:		total length of the first packet for the
 *			aggregation.
 * @pkt_len:		length of all segments
 *
 * Approximate value of the MSS for this aggregation calculated using
 * the first packet of it.
 * Compute number of aggregated segments, and gso_type.
 */
static void bnx2x_set_gro_params(struct sk_buff *skb, u16 parsing_flags,
				 u16 len_on_bd, unsigned int pkt_len,
				 u16 num_of_coalesced_segs)
{
	/* TPA aggregation won't have either IP options or TCP options
	 * other than timestamp or IPv6 extension headers.
	 */
	u16 hdrs_len = ETH_HLEN + sizeof(struct tcphdr);

	if (GET_FLAG(parsing_flags, PARSING_FLAGS_OVER_ETHERNET_PROTOCOL) ==
	    PRS_FLAG_OVERETH_IPV6) {
		hdrs_len += sizeof(struct ipv6hdr);
		skb_shinfo(skb)->gso_type = SKB_GSO_TCPV6;
	} else {
		hdrs_len += sizeof(struct iphdr);
		skb_shinfo(skb)->gso_type = SKB_GSO_TCPV4;
	}

	/* Check if there was a TCP timestamp, if there is it's will
	 * always be 12 bytes length: nop nop kind length echo val.
	 *
	 * Otherwise FW would close the aggregation.
	 */
	if (parsing_flags & PARSING_FLAGS_TIME_STAMP_EXIST_FLAG)
		hdrs_len += TPA_TSTAMP_OPT_LEN;

	skb_shinfo(skb)->gso_size = len_on_bd - hdrs_len;

	/* tcp_gro_complete() will copy NAPI_GRO_CB(skb)->count
	 * to skb_shinfo(skb)->gso_segs
	 */
	NAPI_GRO_CB(skb)->count = num_of_coalesced_segs;
}

static int bnx2x_alloc_rx_sge(struct bnx2x *bp, struct bnx2x_fastpath *fp,
			      u16 index, gfp_t gfp_mask)
{
	struct sw_rx_page *sw_buf = &fp->rx_page_ring[index];
	struct eth_rx_sge *sge = &fp->rx_sge_ring[index];
	struct bnx2x_alloc_pool *pool = &fp->page_pool;
	dma_addr_t mapping;

	if (!pool->page || (PAGE_SIZE - pool->offset) < SGE_PAGE_SIZE) {

		/* put page reference used by the memory pool, since we
		 * won't be using this page as the mempool anymore.
		 */
		if (pool->page)
			put_page(pool->page);

		pool->page = alloc_pages(gfp_mask, PAGES_PER_SGE_SHIFT);
		if (unlikely(!pool->page))
			return -ENOMEM;

		pool->offset = 0;
	}

	mapping = dma_map_page(&bp->pdev->dev, pool->page,
			       pool->offset, SGE_PAGE_SIZE, DMA_FROM_DEVICE);
	if (unlikely(dma_mapping_error(&bp->pdev->dev, mapping))) {
		BNX2X_ERR("Can't map sge\n");
		return -ENOMEM;
	}

	get_page(pool->page);
	sw_buf->page = pool->page;
	sw_buf->offset = pool->offset;

	dma_unmap_addr_set(sw_buf, mapping, mapping);

	sge->addr_hi = cpu_to_le32(U64_HI(mapping));
	sge->addr_lo = cpu_to_le32(U64_LO(mapping));

	pool->offset += SGE_PAGE_SIZE;

	return 0;
}

static int bnx2x_fill_frag_skb(struct bnx2x *bp, struct bnx2x_fastpath *fp,
			       struct bnx2x_agg_info *tpa_info,
			       u16 pages,
			       struct sk_buff *skb,
			       struct eth_end_agg_rx_cqe *cqe,
			       u16 cqe_idx)
{
	struct sw_rx_page *rx_pg, old_rx_pg;
	u32 i, frag_len, frag_size;
	int err, j, frag_id = 0;
	u16 len_on_bd = tpa_info->len_on_bd;
	u16 full_page = 0, gro_size = 0;

	frag_size = le16_to_cpu(cqe->pkt_len) - len_on_bd;

	if (fp->mode == TPA_MODE_GRO) {
		gro_size = tpa_info->gro_size;
		full_page = tpa_info->full_page;
	}

	/* This is needed in order to enable forwarding support */
	if (frag_size)
		bnx2x_set_gro_params(skb, tpa_info->parsing_flags, len_on_bd,
				     le16_to_cpu(cqe->pkt_len),
				     le16_to_cpu(cqe->num_of_coalesced_segs));

#ifdef BNX2X_STOP_ON_ERROR
	if (pages > min_t(u32, 8, MAX_SKB_FRAGS) * SGE_PAGES) {
		BNX2X_ERR("SGL length is too long: %d. CQE index is %d\n",
			  pages, cqe_idx);
		BNX2X_ERR("cqe->pkt_len = %d\n", cqe->pkt_len);
		bnx2x_panic();
		return -EINVAL;
	}
#endif

	/* Run through the SGL and compose the fragmented skb */
	for (i = 0, j = 0; i < pages; i += PAGES_PER_SGE, j++) {
		u16 sge_idx = RX_SGE(le16_to_cpu(cqe->sgl_or_raw_data.sgl[j]));

		/* FW gives the indices of the SGE as if the ring is an array
		   (meaning that "next" element will consume 2 indices) */
		if (fp->mode == TPA_MODE_GRO)
			frag_len = min_t(u32, frag_size, (u32)full_page);
		else /* LRO */
			frag_len = min_t(u32, frag_size, (u32)SGE_PAGES);

		rx_pg = &fp->rx_page_ring[sge_idx];
		old_rx_pg = *rx_pg;

		/* If we fail to allocate a substitute page, we simply stop
		   where we are and drop the whole packet */
		err = bnx2x_alloc_rx_sge(bp, fp, sge_idx, GFP_ATOMIC);
		if (unlikely(err)) {
			bnx2x_fp_qstats(bp, fp)->rx_skb_alloc_failed++;
			return err;
		}

		dma_unmap_page(&bp->pdev->dev,
			       dma_unmap_addr(&old_rx_pg, mapping),
			       SGE_PAGE_SIZE, DMA_FROM_DEVICE);
		/* Add one frag and update the appropriate fields in the skb */
		if (fp->mode == TPA_MODE_LRO)
			skb_fill_page_desc(skb, j, old_rx_pg.page,
					   old_rx_pg.offset, frag_len);
		else { /* GRO */
			int rem;
			int offset = 0;
			for (rem = frag_len; rem > 0; rem -= gro_size) {
				int len = rem > gro_size ? gro_size : rem;
				skb_fill_page_desc(skb, frag_id++,
						   old_rx_pg.page,
						   old_rx_pg.offset + offset,
						   len);
				if (offset)
					get_page(old_rx_pg.page);
				offset += len;
			}
		}

		skb->data_len += frag_len;
		skb->truesize += SGE_PAGES;
		skb->len += frag_len;

		frag_size -= frag_len;
	}

	return 0;
}

static void bnx2x_frag_free(const struct bnx2x_fastpath *fp, void *data)
{
	if (fp->rx_frag_size)
		skb_free_frag(data);
	else
		kfree(data);
}

static void *bnx2x_frag_alloc(const struct bnx2x_fastpath *fp, gfp_t gfp_mask)
{
	if (fp->rx_frag_size) {
		/* GFP_KERNEL allocations are used only during initialization */
		if (unlikely(gfpflags_allow_blocking(gfp_mask)))
			return (void *)__get_free_page(gfp_mask);

		return netdev_alloc_frag(fp->rx_frag_size);
	}

	return kmalloc(fp->rx_buf_size + NET_SKB_PAD, gfp_mask);
}

#ifdef CONFIG_INET
static void bnx2x_gro_ip_csum(struct bnx2x *bp, struct sk_buff *skb)
{
	const struct iphdr *iph = ip_hdr(skb);
	struct tcphdr *th;

	skb_set_transport_header(skb, sizeof(struct iphdr));
	th = tcp_hdr(skb);

	th->check = ~tcp_v4_check(skb->len - skb_transport_offset(skb),
				  iph->saddr, iph->daddr, 0);
}

static void bnx2x_gro_ipv6_csum(struct bnx2x *bp, struct sk_buff *skb)
{
	struct ipv6hdr *iph = ipv6_hdr(skb);
	struct tcphdr *th;

	skb_set_transport_header(skb, sizeof(struct ipv6hdr));
	th = tcp_hdr(skb);

	th->check = ~tcp_v6_check(skb->len - skb_transport_offset(skb),
				  &iph->saddr, &iph->daddr, 0);
}

static void bnx2x_gro_csum(struct bnx2x *bp, struct sk_buff *skb,
			    void (*gro_func)(struct bnx2x*, struct sk_buff*))
{
	skb_set_network_header(skb, 0);
	gro_func(bp, skb);
	tcp_gro_complete(skb);
}
#endif

static void bnx2x_gro_receive(struct bnx2x *bp, struct bnx2x_fastpath *fp,
			       struct sk_buff *skb)
{
#ifdef CONFIG_INET
	if (skb_shinfo(skb)->gso_size) {
		switch (be16_to_cpu(skb->protocol)) {
		case ETH_P_IP:
			bnx2x_gro_csum(bp, skb, bnx2x_gro_ip_csum);
			break;
		case ETH_P_IPV6:
			bnx2x_gro_csum(bp, skb, bnx2x_gro_ipv6_csum);
			break;
		default:
			WARN_ONCE(1, "Error: FW GRO supports only IPv4/IPv6, not 0x%04x\n",
				  be16_to_cpu(skb->protocol));
		}
	}
#endif
	skb_record_rx_queue(skb, fp->rx_queue);
	napi_gro_receive(&fp->napi, skb);
}

static void bnx2x_tpa_stop(struct bnx2x *bp, struct bnx2x_fastpath *fp,
			   struct bnx2x_agg_info *tpa_info,
			   u16 pages,
			   struct eth_end_agg_rx_cqe *cqe,
			   u16 cqe_idx)
{
	struct sw_rx_bd *rx_buf = &tpa_info->first_buf;
	u8 pad = tpa_info->placement_offset;
	u16 len = tpa_info->len_on_bd;
	struct sk_buff *skb = NULL;
	u8 *new_data, *data = rx_buf->data;
	u8 old_tpa_state = tpa_info->tpa_state;
#ifdef CONFIG_XEN_SME
	char dbg_buff[DBG_BUF_SIZE];
	memset(dbg_buff, 0, DBG_BUF_SIZE);
#endif

	tpa_info->tpa_state = BNX2X_TPA_STOP;

	/* If we there was an error during the handling of the TPA_START -
	 * drop this aggregation.
	 */
	if (old_tpa_state == BNX2X_TPA_ERROR)
		goto drop;

	/* Try to allocate the new data */
	new_data = bnx2x_frag_alloc(fp, GFP_ATOMIC);
	/* Unmap skb in the pool anyway, as we are going to change
	   pool entry status to BNX2X_TPA_STOP even if new skb allocation
	   fails. */
	dma_unmap_single(&bp->pdev->dev, dma_unmap_addr(rx_buf, mapping),
			 fp->rx_buf_size, DMA_FROM_DEVICE);
	if (likely(new_data))
		skb = build_skb(data, fp->rx_frag_size);

	if (likely(skb)) {
#ifdef BNX2X_STOP_ON_ERROR
		if (pad + len > fp->rx_buf_size) {
			BNX2X_ERR("skb_put is about to fail...  pad %d  len %d  rx_buf_size %d\n",
				  pad, len, fp->rx_buf_size);
			bnx2x_panic();
			return;
		}
#endif

		skb_reserve(skb, pad + NET_SKB_PAD);
		skb_put(skb, len);
		skb_set_hash(skb, tpa_info->rxhash, tpa_info->rxhash_type);

		skb->protocol = eth_type_trans(skb, bp->dev);
		skb->ip_summed = CHECKSUM_UNNECESSARY;

#ifdef CONFIG_XEN_SME
		memset(dbg_buff, 0, DBG_BUF_SIZE);
		sprintf(dbg_buff, "%s:%d caller: %pS", __func__, __LINE__, __builtin_return_address(0));
		xsl_print_sk_buff(skb, dbg_buff);
#endif
		if (!bnx2x_fill_frag_skb(bp, fp, tpa_info, pages,
					 skb, cqe, cqe_idx)) {
			if (tpa_info->parsing_flags & PARSING_FLAGS_VLAN)
				__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), tpa_info->vlan_tag);
			bnx2x_gro_receive(bp, fp, skb);
		} else {
			DP(NETIF_MSG_RX_STATUS,
			   "Failed to allocate new pages - dropping packet!\n");
			dev_kfree_skb_any(skb);
		}

		/* put new data in bin */
		rx_buf->data = new_data;

		return;
	}
	if (new_data)
		bnx2x_frag_free(fp, new_data);
drop:
	/* drop the packet and keep the buffer in the bin */
	DP(NETIF_MSG_RX_STATUS,
	   "Failed to allocate or map a new skb - dropping packet!\n");
	bnx2x_fp_stats(bp, fp)->eth_q_stats.rx_skb_alloc_failed++;
}

static int bnx2x_alloc_rx_data(struct bnx2x *bp, struct bnx2x_fastpath *fp,
			       u16 index, gfp_t gfp_mask)
{
	u8 *data;
	struct sw_rx_bd *rx_buf = &fp->rx_buf_ring[index];
	struct eth_rx_bd *rx_bd = &fp->rx_desc_ring[index];
	dma_addr_t mapping;

	data = bnx2x_frag_alloc(fp, gfp_mask);
	if (unlikely(data == NULL))
		return -ENOMEM;

	mapping = dma_map_single(&bp->pdev->dev, data + NET_SKB_PAD,
				 fp->rx_buf_size,
				 DMA_FROM_DEVICE);
	if (unlikely(dma_mapping_error(&bp->pdev->dev, mapping))) {
		bnx2x_frag_free(fp, data);
		BNX2X_ERR("Can't map rx data\n");
		return -ENOMEM;
	}

	rx_buf->data = data;
	dma_unmap_addr_set(rx_buf, mapping, mapping);

	rx_bd->addr_hi = cpu_to_le32(U64_HI(mapping));
	rx_bd->addr_lo = cpu_to_le32(U64_LO(mapping));

	return 0;
}

static
void bnx2x_csum_validate(struct sk_buff *skb, union eth_rx_cqe *cqe,
				 struct bnx2x_fastpath *fp,
				 struct bnx2x_eth_q_stats *qstats)
{
	/* Do nothing if no L4 csum validation was done.
	 * We do not check whether IP csum was validated. For IPv4 we assume
	 * that if the card got as far as validating the L4 csum, it also
	 * validated the IP csum. IPv6 has no IP csum.
	 */
	if (cqe->fast_path_cqe.status_flags &
	    ETH_FAST_PATH_RX_CQE_L4_XSUM_NO_VALIDATION_FLG)
		return;

	/* If L4 validation was done, check if an error was found. */

	if (cqe->fast_path_cqe.type_error_flags &
	    (ETH_FAST_PATH_RX_CQE_IP_BAD_XSUM_FLG |
	     ETH_FAST_PATH_RX_CQE_L4_BAD_XSUM_FLG))
		qstats->hw_csum_err++;
	else
		skb->ip_summed = CHECKSUM_UNNECESSARY;
}

static int __bnx2x_rx_int(struct bnx2x_fastpath *fp, int budget)
{
	struct bnx2x *bp = fp->bp;
	u16 bd_cons, bd_prod, bd_prod_fw, comp_ring_cons;
	u16 sw_comp_cons, sw_comp_prod;
	int rx_pkt = 0;
	union eth_rx_cqe *cqe;
	struct eth_fast_path_rx_cqe *cqe_fp;

#ifdef CONFIG_XEN_SME
	char dbg_buff[DBG_BUF_SIZE];
	u16 bd_cons_fw;
#endif

	struct sw_rx_bd *rx_buf = NULL;
	struct sk_buff *skb = NULL;
	u8 cqe_fp_flags;
	enum eth_rx_cqe_type cqe_fp_type;
	u16 len, pad, queue;
	u8 *data;
	u32 rxhash;
	enum pkt_hash_types rxhash_type;

#ifdef CONFIG_XEN_SME
	if (lnk_intercept_rx_path) {
		bd_cons = fp->swi_bd_cons;
		bd_prod = fp->swi_bd_prod;
		sw_comp_cons = fp->swi_comp_cons;
		sw_comp_prod = fp->swi_comp_prod;
	} else {
#endif
		bd_cons = fp->rx_bd_cons;
		bd_prod = fp->rx_bd_prod;
		sw_comp_cons = fp->rx_comp_cons;
		sw_comp_prod = fp->rx_comp_prod;
#ifdef CONFIG_XEN_SME
	}
	bd_cons_fw = bd_cons;
#endif
	
	bd_prod_fw = bd_prod;

	comp_ring_cons = RCQ_BD(sw_comp_cons);
	cqe = &fp->rx_comp_ring[comp_ring_cons];
	cqe_fp = &cqe->fast_path_cqe;

#ifdef BNX2X_STOP_ON_ERROR
	if (unlikely(bp->panic))
		return -5;
#endif

	bd_prod = RX_BD(bd_prod);
	bd_cons = RX_BD(bd_cons);

	/* A rmb() is required to ensure that the CQE is not read
	 * before it is written by the adapter DMA.  PCI ordering
	 * rules will make sure the other fields are written before
	 * the marker at the end of struct eth_fast_path_rx_cqe
	 * but without rmb() a weakly ordered processor can process
	 * stale data.  Without the barrier TPA state-machine might
	 * enter inconsistent state and kernel stack might be
	 * provided with incorrect packet description - these lead
	 * to various kernel crashed.
	 */
	rmb();

	cqe_fp_flags = cqe_fp->type_error_flags;
	cqe_fp_type = cqe_fp_flags & ETH_FAST_PATH_RX_CQE_TYPE;

	DP(NETIF_MSG_RX_STATUS,
		 "CQE type %x  err %x  status %x  queue %x  vlan %x  len %u\n",
		 CQE_TYPE(cqe_fp_flags),
		 cqe_fp_flags, cqe_fp->status_flags,
		 le32_to_cpu(cqe_fp->rss_hash_result),
		 le16_to_cpu(cqe_fp->vlan_tag),
		 le16_to_cpu(cqe_fp->pkt_len_or_gro_seg_len));

	/* is this a slowpath msg? */
	if (unlikely(CQE_TYPE_SLOW(cqe_fp_type))) {
		bnx2x_sp_event(fp, cqe);
		goto next_cqe;
	}

	rx_buf = &fp->rx_buf_ring[bd_cons];
	data = rx_buf->data;

	if (!CQE_TYPE_FAST(cqe_fp_type)) {
		struct bnx2x_agg_info *tpa_info;
		u16 frag_size, pages;
#ifdef BNX2X_STOP_ON_ERROR
		/* sanity check */
		if (fp->mode == TPA_MODE_DISABLED &&
				(CQE_TYPE_START(cqe_fp_type) ||
				 CQE_TYPE_STOP(cqe_fp_type)))
			BNX2X_ERR("START/STOP packet while TPA disabled, type %x\n",
					CQE_TYPE(cqe_fp_type));