ixgbe: Add infrastructure code for FCoE large receive offload to 82599

This adds infrastructure code for FCoE Rx side offload feature to
82599, which provides large receive offload for FCoE by Direct
Data Placement (DDP). The ixgbe_fcoe_ddp_get() and ixgbe_fcoe_ddp_put()
pair corresponds to the netdev support to FCoE by the function pointers
provided in net_device_ops as ndo_fcoe_ddp_setup and ndo_fcoe_ddp_done.
The implementation of these in ixgbe is shown in the next patch.

Signed-off-by: Yi Zou <yi.zou@intel.com>
Signed-off-by: Chris Leech <christopher.leech@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

3 files changed, 344 insertions, 1 deletions

diff --git a/drivers/net/ixgbe/ixgbe.h b/drivers/net/ixgbe/ixgbe.h
index 06859f687f19..5581fa3976df 100644
--- a/drivers/net/ixgbe/ixgbe.h
+++ b/drivers/net/ixgbe/ixgbe.h
@@ -336,6 +336,9 @@ struct ixgbe_adapter {
 	struct timer_list sfp_timer;
 	struct work_struct multispeed_fiber_task;
 	struct work_struct sfp_config_module_task;
+#ifdef IXGBE_FCOE
+	struct ixgbe_fcoe fcoe;
+#endif /* IXGBE_FCOE */
 	u64 rsc_count;
 	u32 wol;
 	u16 eeprom_version;
diff --git a/drivers/net/ixgbe/ixgbe_fcoe.c b/drivers/net/ixgbe/ixgbe_fcoe.c
index e20facd1f2df..a11c6207b748 100644
--- a/drivers/net/ixgbe/ixgbe_fcoe.c
+++ b/drivers/net/ixgbe/ixgbe_fcoe.c
@@ -36,6 +36,311 @@
 #include <scsi/libfcoe.h>
 
 /**
+ * ixgbe_rx_is_fcoe - check the rx desc for incoming pkt type
+ * @rx_desc: advanced rx descriptor
+ *
+ * Returns : true if it is FCoE pkt
+ */
+static inline bool ixgbe_rx_is_fcoe(union ixgbe_adv_rx_desc *rx_desc)
+{
+	u16 p;
+
+	p = le16_to_cpu(rx_desc->wb.lower.lo_dword.hs_rss.pkt_info);
+	if (p & IXGBE_RXDADV_PKTTYPE_ETQF) {
+		p &= IXGBE_RXDADV_PKTTYPE_ETQF_MASK;
+		p >>= IXGBE_RXDADV_PKTTYPE_ETQF_SHIFT;
+		return p == IXGBE_ETQF_FILTER_FCOE;
+	}
+	return false;
+}
+
+/**
+ * ixgbe_fcoe_clear_ddp - clear the given ddp context
+ * @ddp - ptr to the ixgbe_fcoe_ddp
+ *
+ * Returns : none
+ *
+ */
+static inline void ixgbe_fcoe_clear_ddp(struct ixgbe_fcoe_ddp *ddp)
+{
+	ddp->len = 0;
+	ddp->err = 0;
+	ddp->udl = NULL;
+	ddp->udp = 0UL;
+	ddp->sgl = NULL;
+	ddp->sgc = 0;
+}
+
+/**
+ * ixgbe_fcoe_ddp_put - free the ddp context for a given xid
+ * @netdev: the corresponding net_device
+ * @xid: the xid that corresponding ddp will be freed
+ *
+ * This is the implementation of net_device_ops.ndo_fcoe_ddp_done
+ * and it is expected to be called by ULD, i.e., FCP layer of libfc
+ * to release the corresponding ddp context when the I/O is done.
+ *
+ * Returns : data length already ddp-ed in bytes
+ */
+int ixgbe_fcoe_ddp_put(struct net_device *netdev, u16 xid)
+{
+	int len = 0;
+	struct ixgbe_fcoe *fcoe;
+	struct ixgbe_adapter *adapter;
+	struct ixgbe_fcoe_ddp *ddp;
+
+	if (!netdev)
+		goto out_ddp_put;
+
+	if (xid >= IXGBE_FCOE_DDP_MAX)
+		goto out_ddp_put;
+
+	adapter = netdev_priv(netdev);
+	fcoe = &adapter->fcoe;
+	ddp = &fcoe->ddp[xid];
+	if (!ddp->udl)
+		goto out_ddp_put;
+
+	len = ddp->len;
+	/* if there an error, force to invalidate ddp context */
+	if (ddp->err) {
+		spin_lock_bh(&fcoe->lock);
+		IXGBE_WRITE_REG(&adapter->hw, IXGBE_FCFLT, 0);
+		IXGBE_WRITE_REG(&adapter->hw, IXGBE_FCFLTRW,
+				(xid | IXGBE_FCFLTRW_WE));
+		IXGBE_WRITE_REG(&adapter->hw, IXGBE_FCBUFF, 0);
+		IXGBE_WRITE_REG(&adapter->hw, IXGBE_FCDMARW,
+				(xid | IXGBE_FCDMARW_WE));
+		spin_unlock_bh(&fcoe->lock);
+	}
+	if (ddp->sgl)
+		pci_unmap_sg(adapter->pdev, ddp->sgl, ddp->sgc,
+			     DMA_FROM_DEVICE);
+	pci_pool_free(fcoe->pool, ddp->udl, ddp->udp);
+	ixgbe_fcoe_clear_ddp(ddp);
+
+out_ddp_put:
+	return len;
+}
+
+/**
+ * ixgbe_fcoe_ddp_get - called to set up ddp context
+ * @netdev: the corresponding net_device
+ * @xid: the exchange id requesting ddp
+ * @sgl: the scatter-gather list for this request
+ * @sgc: the number of scatter-gather items
+ *
+ * This is the implementation of net_device_ops.ndo_fcoe_ddp_setup
+ * and is expected to be called from ULD, e.g., FCP layer of libfc
+ * to set up ddp for the corresponding xid of the given sglist for
+ * the corresponding I/O.
+ *
+ * Returns : 1 for success and 0 for no ddp
+ */
+int ixgbe_fcoe_ddp_get(struct net_device *netdev, u16 xid,
+		       struct scatterlist *sgl, unsigned int sgc)
+{
+	struct ixgbe_adapter *adapter;
+	struct ixgbe_hw *hw;
+	struct ixgbe_fcoe *fcoe;
+	struct ixgbe_fcoe_ddp *ddp;
+	struct scatterlist *sg;
+	unsigned int i, j, dmacount;
+	unsigned int len;
+	static const unsigned int bufflen = 4096;
+	unsigned int firstoff = 0;
+	unsigned int lastsize;
+	unsigned int thisoff = 0;
+	unsigned int thislen = 0;
+	u32 fcbuff, fcdmarw, fcfltrw;
+	dma_addr_t addr;
+
+	if (!netdev || !sgl)
+		return 0;
+
+	adapter = netdev_priv(netdev);
+	if (xid >= IXGBE_FCOE_DDP_MAX) {
+		DPRINTK(DRV, WARNING, "xid=0x%x out-of-range\n", xid);
+		return 0;
+	}
+
+	fcoe = &adapter->fcoe;
+	if (!fcoe->pool) {
+		DPRINTK(DRV, WARNING, "xid=0x%x no ddp pool for fcoe\n", xid);
+		return 0;
+	}
+
+	ddp = &fcoe->ddp[xid];
+	if (ddp->sgl) {
+		DPRINTK(DRV, ERR, "xid 0x%x w/ non-null sgl=%p nents=%d\n",
+			xid, ddp->sgl, ddp->sgc);
+		return 0;
+	}
+	ixgbe_fcoe_clear_ddp(ddp);
+
+	/* setup dma from scsi command sgl */
+	dmacount = pci_map_sg(adapter->pdev, sgl, sgc, DMA_FROM_DEVICE);
+	if (dmacount == 0) {
+		DPRINTK(DRV, ERR, "xid 0x%x DMA map error\n", xid);
+		return 0;
+	}
+
+	/* alloc the udl from our ddp pool */
+	ddp->udl = pci_pool_alloc(fcoe->pool, GFP_KERNEL, &ddp->udp);
+	if (!ddp->udl) {
+		DPRINTK(DRV, ERR, "failed allocated ddp context\n");
+		goto out_noddp_unmap;
+	}
+	ddp->sgl = sgl;
+	ddp->sgc = sgc;
+
+	j = 0;
+	for_each_sg(sgl, sg, dmacount, i) {
+		addr = sg_dma_address(sg);
+		len = sg_dma_len(sg);
+		while (len) {
+			/* get the offset of length of current buffer */
+			thisoff = addr & ((dma_addr_t)bufflen - 1);
+			thislen = min((bufflen - thisoff), len);
+			/*
+			 * all but the 1st buffer (j == 0)
+			 * must be aligned on bufflen
+			 */
+			if ((j != 0) && (thisoff))
+				goto out_noddp_free;
+			/*
+			 * all but the last buffer
+			 * ((i == (dmacount - 1)) && (thislen == len))
+			 * must end at bufflen
+			 */
+			if (((i != (dmacount - 1)) || (thislen != len))
+			    && ((thislen + thisoff) != bufflen))
+				goto out_noddp_free;
+
+			ddp->udl[j] = (u64)(addr - thisoff);
+			/* only the first buffer may have none-zero offset */
+			if (j == 0)
+				firstoff = thisoff;
+			len -= thislen;
+			addr += thislen;
+			j++;
+			/* max number of buffers allowed in one DDP context */
+			if (j > IXGBE_BUFFCNT_MAX) {
+				DPRINTK(DRV, ERR, "xid=%x:%d,%d,%d:addr=%llx "
+					"not enough descriptors\n",
+					xid, i, j, dmacount, (u64)addr);
+				goto out_noddp_free;
+			}
+		}
+	}
+	/* only the last buffer may have non-full bufflen */
+	lastsize = thisoff + thislen;
+
+	fcbuff = (IXGBE_FCBUFF_4KB << IXGBE_FCBUFF_BUFFSIZE_SHIFT);
+	fcbuff |= (j << IXGBE_FCBUFF_BUFFCNT_SHIFT);
+	fcbuff |= (firstoff << IXGBE_FCBUFF_OFFSET_SHIFT);
+	fcbuff |= (IXGBE_FCBUFF_VALID);
+
+	fcdmarw = xid;
+	fcdmarw |= IXGBE_FCDMARW_WE;
+	fcdmarw |= (lastsize << IXGBE_FCDMARW_LASTSIZE_SHIFT);
+
+	fcfltrw = xid;
+	fcfltrw |= IXGBE_FCFLTRW_WE;
+
+	/* program DMA context */
+	hw = &adapter->hw;
+	spin_lock_bh(&fcoe->lock);
+	IXGBE_WRITE_REG(hw, IXGBE_FCPTRL, ddp->udp & DMA_32BIT_MASK);
+	IXGBE_WRITE_REG(hw, IXGBE_FCPTRH, (u64)ddp->udp >> 32);
+	IXGBE_WRITE_REG(hw, IXGBE_FCBUFF, fcbuff);
+	IXGBE_WRITE_REG(hw, IXGBE_FCDMARW, fcdmarw);
+	/* program filter context */
+	IXGBE_WRITE_REG(hw, IXGBE_FCPARAM, 0);
+	IXGBE_WRITE_REG(hw, IXGBE_FCFLT, IXGBE_FCFLT_VALID);
+	IXGBE_WRITE_REG(hw, IXGBE_FCFLTRW, fcfltrw);
+	spin_unlock_bh(&fcoe->lock);
+
+	return 1;
+
+out_noddp_free:
+	pci_pool_free(fcoe->pool, ddp->udl, ddp->udp);
+	ixgbe_fcoe_clear_ddp(ddp);
+
+out_noddp_unmap:
+	pci_unmap_sg(adapter->pdev, sgl, sgc, DMA_FROM_DEVICE);
+	return 0;
+}
+
+/**
+ * ixgbe_fcoe_ddp - check ddp status and mark it done
+ * @adapter: ixgbe adapter
+ * @rx_desc: advanced rx descriptor
+ * @skb: the skb holding the received data
+ *
+ * This checks ddp status.
+ *
+ * Returns : 0 for success and skb will not be delivered to ULD
+ */
+int ixgbe_fcoe_ddp(struct ixgbe_adapter *adapter,
+		   union ixgbe_adv_rx_desc *rx_desc,
+		   struct sk_buff *skb)
+{
+	u16 xid;
+	u32 sterr, fceofe, fcerr, fcstat;
+	int rc = -EINVAL;
+	struct ixgbe_fcoe *fcoe;
+	struct ixgbe_fcoe_ddp *ddp;
+	struct fc_frame_header *fh;
+
+	if (!ixgbe_rx_is_fcoe(rx_desc))
+		goto ddp_out;
+
+	skb->ip_summed = CHECKSUM_UNNECESSARY;
+	sterr = le32_to_cpu(rx_desc->wb.upper.status_error);
+	fcerr = (sterr & IXGBE_RXDADV_ERR_FCERR);
+	fceofe = (sterr & IXGBE_RXDADV_ERR_FCEOFE);
+	if (fcerr == IXGBE_FCERR_BADCRC)
+		skb->ip_summed = CHECKSUM_NONE;
+
+	skb_reset_network_header(skb);
+	skb_set_transport_header(skb, skb_network_offset(skb) +
+				 sizeof(struct fcoe_hdr));
+	fh = (struct fc_frame_header *)skb_transport_header(skb);
+	xid =  be16_to_cpu(fh->fh_ox_id);
+	if (xid >= IXGBE_FCOE_DDP_MAX)
+		goto ddp_out;
+
+	fcoe = &adapter->fcoe;
+	ddp = &fcoe->ddp[xid];
+	if (!ddp->udl)
+		goto ddp_out;
+
+	ddp->err = (fcerr | fceofe);
+	if (ddp->err)
+		goto ddp_out;
+
+	fcstat = (sterr & IXGBE_RXDADV_STAT_FCSTAT);
+	if (fcstat) {
+		/* update length of DDPed data */
+		ddp->len = le32_to_cpu(rx_desc->wb.lower.hi_dword.rss);
+		/* unmap the sg list when FCP_RSP is received */
+		if (fcstat == IXGBE_RXDADV_STAT_FCSTAT_FCPRSP) {
+			pci_unmap_sg(adapter->pdev, ddp->sgl,
+				     ddp->sgc, DMA_FROM_DEVICE);
+			ddp->sgl = NULL;
+			ddp->sgc = 0;
+		}
+		/* return 0 to bypass going to ULD for DDPed data */
+		if (fcstat == IXGBE_RXDADV_STAT_FCSTAT_DDP)
+			rc = 0;
+	}
+
+ddp_out:
+	return rc;
+}
+
+/**
  * ixgbe_fso - ixgbe FCoE Sequence Offload (FSO)
  * @adapter: ixgbe adapter
  * @tx_ring: tx desc ring
@@ -178,7 +483,20 @@ int ixgbe_fso(struct ixgbe_adapter *adapter,
 void ixgbe_configure_fcoe(struct ixgbe_adapter *adapter)
 {
 	struct ixgbe_hw *hw = &adapter->hw;
+	struct ixgbe_fcoe *fcoe = &adapter->fcoe;
+
+	/* create the pool for ddp if not created yet */
+	if (!fcoe->pool) {
+		/* allocate ddp pool */
+		fcoe->pool = pci_pool_create("ixgbe_fcoe_ddp",
+					     adapter->pdev, IXGBE_FCPTR_MAX,
+					     IXGBE_FCPTR_ALIGN, PAGE_SIZE);
+		if (!fcoe->pool)
+			DPRINTK(DRV, ERR,
+				"failed to allocated FCoE DDP pool\n");
 
+		spin_lock_init(&fcoe->lock);
+	}
 	/* L2 filter for FCoE: default to queue 0 */
 	IXGBE_WRITE_REG(hw, IXGBE_ETQF(IXGBE_ETQF_FILTER_FCOE),
 			(ETH_P_FCOE | IXGBE_ETQF_FCOE | IXGBE_ETQF_FILTER_EN));
@@ -190,3 +508,25 @@ void ixgbe_configure_fcoe(struct ixgbe_adapter *adapter)
 			IXGBE_FCRXCTRL_FCCRCBO |
 			(FC_FCOE_VER << IXGBE_FCRXCTRL_FCOEVER_SHIFT));
 }
+
+/**
+ * ixgbe_cleanup_fcoe - release all fcoe ddp context resources
+ * @adapter : ixgbe adapter
+ *
+ * Cleans up outstanding ddp context resources
+ *
+ * Returns : none
+ */
+void ixgbe_cleanup_fcoe(struct ixgbe_adapter *adapter)
+{
+	int i;
+	struct ixgbe_fcoe *fcoe = &adapter->fcoe;
+
+	/* release ddp resource */
+	if (fcoe->pool) {
+		for (i = 0; i < IXGBE_FCOE_DDP_MAX; i++)
+			ixgbe_fcoe_ddp_put(adapter->netdev, i);
+		pci_pool_destroy(fcoe->pool);
+		fcoe->pool = NULL;
+	}
+}
diff --git a/drivers/net/ixgbe/ixgbe_fcoe.h b/drivers/net/ixgbe/ixgbe_fcoe.h
index d054c1dada50..b7f9b63aa49f 100644
--- a/drivers/net/ixgbe/ixgbe_fcoe.h
+++ b/drivers/net/ixgbe/ixgbe_fcoe.h
@@ -54,7 +54,7 @@ struct ixgbe_fcoe_ddp {
 	unsigned int sgc;
 	struct scatterlist *sgl;
 	dma_addr_t udp;
-	unsigned long *udl;
+	u64 *udl;
 };
 
 struct ixgbe_fcoe {