ixgbe: add LRO support

Support for in-kernel LRO with the ability to enable/disable via ethtool
based on comments from Ben Hutchings.
Signed-off-by: NMallikarjuna R Chilakala <mallikarjuna.chilakala@intel.com>
Signed-off-by: NJeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: NPJ Waskiewicz <peter.p.waskiewicz.jr@intel.com>
Signed-off-by: NJeff Garzik <jgarzik@redhat.com>

drivers/net/Kconfig

@@ -2471,7 +2471,8 @@ config EHEA
 
 config IXGBE
 	tristate "Intel(R) 10GbE PCI Express adapters support"
-	depends on PCI
+	depends on PCI && INET
+	select INET_LRO
 	---help---
 	  This driver supports Intel(R) 10GbE PCI Express family of
 	  adapters.  For more information on how to identify your adapter, go

drivers/net/ixgbe/ixgbe.h

@@ -32,6 +32,7 @@
 #include <linux/types.h>
 #include <linux/pci.h>
 #include <linux/netdevice.h>
+#include <linux/inet_lro.h>
 
 #include "ixgbe_type.h"
 #include "ixgbe_common.h"
@@ -100,6 +101,9 @@
 #define IXGBE_TX_FLAGS_VLAN_MASK	0xffff0000
 #define IXGBE_TX_FLAGS_VLAN_SHIFT	16
 
+#define IXGBE_MAX_LRO_DESCRIPTORS       8
+#define IXGBE_MAX_LRO_AGGREGATE         32
+
 /* wrapper around a pointer to a socket buffer,
  * so a DMA handle can be stored along with the buffer */
 struct ixgbe_tx_buffer {
@@ -150,6 +154,8 @@ struct ixgbe_ring {
 	/* cpu for tx queue */
 	int cpu;
 #endif
+	struct net_lro_mgr lro_mgr;
+	bool lro_used;
 	struct ixgbe_queue_stats stats;
 	u8 v_idx; /* maps directly to the index for this ring in the hardware
 		   * vector array, can also be used for finding the bit in EICR
@@ -287,6 +293,9 @@ struct ixgbe_adapter {
 
 	unsigned long state;
 	u64 tx_busy;
+	u64 lro_aggregated;
+	u64 lro_flushed;
+	u64 lro_no_desc;
 };
 
 enum ixbge_state_t {

drivers/net/ixgbe/ixgbe_ethtool.c

@@ -90,6 +90,8 @@ static struct ixgbe_stats ixgbe_gstrings_stats[] = {
 	{"rx_header_split", IXGBE_STAT(rx_hdr_split)},
 	{"alloc_rx_page_failed", IXGBE_STAT(alloc_rx_page_failed)},
 	{"alloc_rx_buff_failed", IXGBE_STAT(alloc_rx_buff_failed)},
+	{"lro_aggregated", IXGBE_STAT(lro_aggregated)},
+	{"lro_flushed", IXGBE_STAT(lro_flushed)},
 };
 
 #define IXGBE_QUEUE_STATS_LEN \
@@ -787,6 +789,7 @@ static void ixgbe_get_ethtool_stats(struct net_device *netdev,
 	int stat_count = sizeof(struct ixgbe_queue_stats) / sizeof(u64);
 	int j, k;
 	int i;
+	u64 aggregated = 0, flushed = 0, no_desc = 0;
 
 	ixgbe_update_stats(adapter);
 	for (i = 0; i < IXGBE_GLOBAL_STATS_LEN; i++) {
@@ -801,11 +804,17 @@ static void ixgbe_get_ethtool_stats(struct net_device *netdev,
 		i += k;
 	}
 	for (j = 0; j < adapter->num_rx_queues; j++) {
+		aggregated += adapter->rx_ring[j].lro_mgr.stats.aggregated;
+		flushed += adapter->rx_ring[j].lro_mgr.stats.flushed;
+		no_desc += adapter->rx_ring[j].lro_mgr.stats.no_desc;
 		queue_stat = (u64 *)&adapter->rx_ring[j].stats;
 		for (k = 0; k < stat_count; k++)
 			data[i + k] = queue_stat[k];
 		i += k;
 	}
+	adapter->lro_aggregated = aggregated;
+	adapter->lro_flushed = flushed;
+	adapter->lro_no_desc = no_desc;
 }
 
 static void ixgbe_get_strings(struct net_device *netdev, u32 stringset,
@@ -973,6 +982,8 @@ static struct ethtool_ops ixgbe_ethtool_ops = {
 	.get_ethtool_stats      = ixgbe_get_ethtool_stats,
 	.get_coalesce           = ixgbe_get_coalesce,
 	.set_coalesce           = ixgbe_set_coalesce,
+	.get_flags              = ethtool_op_get_flags,
+	.set_flags              = ethtool_op_set_flags,
 };
 
 void ixgbe_set_ethtool_ops(struct net_device *netdev)

drivers/net/ixgbe/ixgbe_main.c

@@ -389,24 +389,39 @@ static int __ixgbe_notify_dca(struct device *dev, void *data)
  * ixgbe_receive_skb - Send a completed packet up the stack
  * @adapter: board private structure
  * @skb: packet to send up
- * @is_vlan: packet has a VLAN tag
- * @tag: VLAN tag from descriptor
+ * @status: hardware indication of status of receive
+ * @rx_ring: rx descriptor ring (for a specific queue) to setup
+ * @rx_desc: rx descriptor
  **/
 static void ixgbe_receive_skb(struct ixgbe_adapter *adapter,
-			      struct sk_buff *skb, bool is_vlan,
-			      u16 tag)
+			      struct sk_buff *skb, u8 status,
+			      struct ixgbe_ring *ring,
+                              union ixgbe_adv_rx_desc *rx_desc)
 {
-	if (!(adapter->flags & IXGBE_FLAG_IN_NETPOLL)) {
-		if (adapter->vlgrp && is_vlan)
-			vlan_hwaccel_receive_skb(skb, adapter->vlgrp, tag);
-		else
-			netif_receive_skb(skb);
-	} else {
+	bool is_vlan = (status & IXGBE_RXD_STAT_VP);
+	u16 tag = le16_to_cpu(rx_desc->wb.upper.vlan);
 
+	if (adapter->netdev->features & NETIF_F_LRO &&
+	    skb->ip_summed == CHECKSUM_UNNECESSARY) {
 		if (adapter->vlgrp && is_vlan)
-			vlan_hwaccel_rx(skb, adapter->vlgrp, tag);
+			lro_vlan_hwaccel_receive_skb(&ring->lro_mgr, skb,
+			                             adapter->vlgrp, tag,
+			                             rx_desc);
 		else
-			netif_rx(skb);
+			lro_receive_skb(&ring->lro_mgr, skb, rx_desc);
+		ring->lro_used = true;
+	} else {
+		if (!(adapter->flags & IXGBE_FLAG_IN_NETPOLL)) {
+			if (adapter->vlgrp && is_vlan)
+				vlan_hwaccel_receive_skb(skb, adapter->vlgrp, tag);
+			else
+				netif_receive_skb(skb);
+		} else {
+			if (adapter->vlgrp && is_vlan)
+				vlan_hwaccel_rx(skb, adapter->vlgrp, tag);
+			else
+				netif_rx(skb);
+		}
 	}
 }
 
@@ -546,8 +561,8 @@ static bool ixgbe_clean_rx_irq(struct ixgbe_adapter *adapter,
 	struct sk_buff *skb;
 	unsigned int i;
 	u32 upper_len, len, staterr;
-	u16 hdr_info, vlan_tag;
-	bool is_vlan, cleaned = false;
+	u16 hdr_info;
+	bool cleaned = false;
 	int cleaned_count = 0;
 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
 
@@ -556,8 +571,6 @@ static bool ixgbe_clean_rx_irq(struct ixgbe_adapter *adapter,
 	rx_desc = IXGBE_RX_DESC_ADV(*rx_ring, i);
 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
 	rx_buffer_info = &rx_ring->rx_buffer_info[i];
-	is_vlan = (staterr & IXGBE_RXD_STAT_VP);
-	vlan_tag = le16_to_cpu(rx_desc->wb.upper.vlan);
 
 	while (staterr & IXGBE_RXD_STAT_DD) {
 		if (*work_done >= work_to_do)
@@ -635,7 +648,7 @@ static bool ixgbe_clean_rx_irq(struct ixgbe_adapter *adapter,
 		total_rx_packets++;
 
 		skb->protocol = eth_type_trans(skb, netdev);
-		ixgbe_receive_skb(adapter, skb, is_vlan, vlan_tag);
+		ixgbe_receive_skb(adapter, skb, staterr, rx_ring, rx_desc);
 		netdev->last_rx = jiffies;
 
 next_desc:
@@ -652,8 +665,11 @@ static bool ixgbe_clean_rx_irq(struct ixgbe_adapter *adapter,
 		rx_buffer_info = next_buffer;
 
 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
-		is_vlan = (staterr & IXGBE_RXD_STAT_VP);
-		vlan_tag = le16_to_cpu(rx_desc->wb.upper.vlan);
+	}
+
+	if (rx_ring->lro_used) {
+		lro_flush_all(&rx_ring->lro_mgr);
+		rx_ring->lro_used = false;
 	}
 
 	rx_ring->next_to_clean = i;
@@ -1381,6 +1397,33 @@ static void ixgbe_configure_tx(struct ixgbe_adapter *adapter)
 			(((S) & (PAGE_SIZE - 1)) ? 1 : 0))
 
 #define IXGBE_SRRCTL_BSIZEHDRSIZE_SHIFT			2
+/**
+ * ixgbe_get_skb_hdr - helper function for LRO header processing
+ * @skb: pointer to sk_buff to be added to LRO packet
+ * @iphdr: pointer to tcp header structure
+ * @tcph: pointer to tcp header structure
+ * @hdr_flags: pointer to header flags
+ * @priv: private data
+ **/
+static int ixgbe_get_skb_hdr(struct sk_buff *skb, void **iphdr, void **tcph,
+                             u64 *hdr_flags, void *priv)
+{
+	union ixgbe_adv_rx_desc *rx_desc = priv;
+
+	/* Verify that this is a valid IPv4 TCP packet */
+	if (!(rx_desc->wb.lower.lo_dword.pkt_info &
+	    (IXGBE_RXDADV_PKTTYPE_IPV4 | IXGBE_RXDADV_PKTTYPE_TCP)))
+		return -1;
+
+	/* Set network headers */
+	skb_reset_network_header(skb);
+	skb_set_transport_header(skb, ip_hdrlen(skb));
+	*iphdr = ip_hdr(skb);
+	*tcph = tcp_hdr(skb);
+	*hdr_flags = LRO_IPV4 | LRO_TCP;
+	return 0;
+}
+
 /**
  * ixgbe_configure_rx - Configure 8254x Receive Unit after Reset
  * @adapter: board private structure
@@ -1470,6 +1513,17 @@ static void ixgbe_configure_rx(struct ixgbe_adapter *adapter)
 		adapter->rx_ring[i].tail = IXGBE_RDT(i);
 	}
 
+	/* Intitial LRO Settings */
+	adapter->rx_ring[i].lro_mgr.max_aggr = IXGBE_MAX_LRO_AGGREGATE;
+	adapter->rx_ring[i].lro_mgr.max_desc = IXGBE_MAX_LRO_DESCRIPTORS;
+	adapter->rx_ring[i].lro_mgr.get_skb_header = ixgbe_get_skb_hdr;
+	adapter->rx_ring[i].lro_mgr.features = LRO_F_EXTRACT_VLAN_ID;
+	if (!(adapter->flags & IXGBE_FLAG_IN_NETPOLL))
+		adapter->rx_ring[i].lro_mgr.features |= LRO_F_NAPI;
+	adapter->rx_ring[i].lro_mgr.dev = adapter->netdev;
+	adapter->rx_ring[i].lro_mgr.ip_summed = CHECKSUM_UNNECESSARY;
+	adapter->rx_ring[i].lro_mgr.ip_summed_aggr = CHECKSUM_UNNECESSARY;
+
 	if (adapter->flags & IXGBE_FLAG_RSS_ENABLED) {
 		/* Fill out redirection table */
 		for (i = 0, j = 0; i < 128; i++, j++) {
@@ -2489,12 +2543,18 @@ int ixgbe_setup_rx_resources(struct ixgbe_adapter *adapter,
 	struct pci_dev *pdev = adapter->pdev;
 	int size;
 
+	size = sizeof(struct net_lro_desc) * IXGBE_MAX_LRO_DESCRIPTORS;
+	rxdr->lro_mgr.lro_arr = vmalloc(size);
+	if (!rxdr->lro_mgr.lro_arr)
+		return -ENOMEM;
+	memset(rxdr->lro_mgr.lro_arr, 0, size);
+
 	size = sizeof(struct ixgbe_rx_buffer) * rxdr->count;
 	rxdr->rx_buffer_info = vmalloc(size);
 	if (!rxdr->rx_buffer_info) {
 		DPRINTK(PROBE, ERR,
 			"vmalloc allocation failed for the rx desc ring\n");
-		return -ENOMEM;
+		goto alloc_failed;
 	}
 	memset(rxdr->rx_buffer_info, 0, size);
 
@@ -2508,13 +2568,18 @@ int ixgbe_setup_rx_resources(struct ixgbe_adapter *adapter,
 		DPRINTK(PROBE, ERR,
 			"Memory allocation failed for the rx desc ring\n");
 		vfree(rxdr->rx_buffer_info);
-		return -ENOMEM;
+		goto alloc_failed;
 	}
 
 	rxdr->next_to_clean = 0;
 	rxdr->next_to_use = 0;
 
 	return 0;
+
+alloc_failed:
+	vfree(rxdr->lro_mgr.lro_arr);
+	rxdr->lro_mgr.lro_arr = NULL;
+	return -ENOMEM;
 }
 
 /**
@@ -2565,6 +2630,9 @@ static void ixgbe_free_rx_resources(struct ixgbe_adapter *adapter,
 {
 	struct pci_dev *pdev = adapter->pdev;
 
+	vfree(rx_ring->lro_mgr.lro_arr);
+	rx_ring->lro_mgr.lro_arr = NULL;
+
 	ixgbe_clean_rx_ring(adapter, rx_ring);
 
 	vfree(rx_ring->rx_buffer_info);
@@ -3517,6 +3585,7 @@ static int __devinit ixgbe_probe(struct pci_dev *pdev,
 			   NETIF_F_HW_VLAN_RX |
 			   NETIF_F_HW_VLAN_FILTER;
 
+	netdev->features |= NETIF_F_LRO;
 	netdev->features |= NETIF_F_TSO;
 	netdev->features |= NETIF_F_TSO6;