[net-next-2.6 PATCH 1/7] igb: change type for ring sizes to u16 in igb_set_ring_param

[net-next-2.6 PATCH 1/7] igb: change type for ring sizes to u16 in igb_set_ring_param

[Jeff Kirsher]

From: Alexander Duyck <alexander.h.duyck@intel.com>

Change the type for the ring size values to u16 and use min/max_t instead of
min/max.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---

 drivers/net/igb/igb_ethtool.c |   10 +++++-----
 1 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/net/igb/igb_ethtool.c b/drivers/net/igb/igb_ethtool.c
index 90b89a8..f2b2825 100644
--- a/drivers/net/igb/igb_ethtool.c
+++ b/drivers/net/igb/igb_ethtool.c
@@ -735,17 +735,17 @@ static int igb_set_ringparam(struct net_device *netdev,
 	struct igb_adapter *adapter = netdev_priv(netdev);
 	struct igb_ring *temp_ring;
 	int i, err = 0;
-	u32 new_rx_count, new_tx_count;
+	u16 new_rx_count, new_tx_count;
 
 	if ((ring->rx_mini_pending) || (ring->rx_jumbo_pending))
 		return -EINVAL;
 
-	new_rx_count = min(ring->rx_pending, (u32)IGB_MAX_RXD);
-	new_rx_count = max(new_rx_count, (u32)IGB_MIN_RXD);
+	new_rx_count = min_t(u32, ring->rx_pending, IGB_MAX_RXD);
+	new_rx_count = max_t(u16, new_rx_count, IGB_MIN_RXD);
 	new_rx_count = ALIGN(new_rx_count, REQ_RX_DESCRIPTOR_MULTIPLE);
 
-	new_tx_count = min(ring->tx_pending, (u32)IGB_MAX_TXD);
-	new_tx_count = max(new_tx_count, (u32)IGB_MIN_TXD);
+	new_tx_count = min_t(u32, ring->tx_pending, IGB_MAX_TXD);
+	new_tx_count = max_t(u16, new_tx_count, IGB_MIN_TXD);
 	new_tx_count = ALIGN(new_tx_count, REQ_TX_DESCRIPTOR_MULTIPLE);
 
 	if ((new_tx_count == adapter->tx_ring_count) &&

[net-next-2.6 PATCH 2/7] igb: move timesync init into a seperate function

[Jeff Kirsher]

From: Alexander Duyck <alexander.h.duyck@intel.com>

Current code is quite large and making igb_probe difficult to read.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---

 drivers/net/igb/igb_main.c |  115 +++++++++++++++++++++++++-------------------
 1 files changed, 65 insertions(+), 50 deletions(-)

diff --git a/drivers/net/igb/igb_main.c b/drivers/net/igb/igb_main.c
index b044c98..d72d484 100644
--- a/drivers/net/igb/igb_main.c
+++ b/drivers/net/igb/igb_main.c
@@ -1566,56 +1566,6 @@ static int __devinit igb_probe(struct pci_dev *pdev,
 	}
 
 #endif
-	switch (hw->mac.type) {
-	case e1000_82576:
-		/*
-		 * Initialize hardware timer: we keep it running just in case
-		 * that some program needs it later on.
-		 */
-		memset(&adapter->cycles, 0, sizeof(adapter->cycles));
-		adapter->cycles.read = igb_read_clock;
-		adapter->cycles.mask = CLOCKSOURCE_MASK(64);
-		adapter->cycles.mult = 1;
-		/**
-		 * Scale the NIC clock cycle by a large factor so that
-		 * relatively small clock corrections can be added or
-		 * substracted at each clock tick. The drawbacks of a large
-		 * factor are a) that the clock register overflows more quickly
-		 * (not such a big deal) and b) that the increment per tick has
-		 * to fit into 24 bits.  As a result we need to use a shift of
-		 * 19 so we can fit a value of 16 into the TIMINCA register.
-		 */
-		adapter->cycles.shift = IGB_82576_TSYNC_SHIFT;
-		wr32(E1000_TIMINCA,
-		                (1 << E1000_TIMINCA_16NS_SHIFT) |
-		                (16 << IGB_82576_TSYNC_SHIFT));
-
-		/* Set registers so that rollover occurs soon to test this. */
-		wr32(E1000_SYSTIML, 0x00000000);
-		wr32(E1000_SYSTIMH, 0xFF800000);
-		wrfl();
-
-		timecounter_init(&adapter->clock,
-				 &adapter->cycles,
-				 ktime_to_ns(ktime_get_real()));
-		/*
-		 * Synchronize our NIC clock against system wall clock. NIC
-		 * time stamp reading requires ~3us per sample, each sample
-		 * was pretty stable even under load => only require 10
-		 * samples for each offset comparison.
-		 */
-		memset(&adapter->compare, 0, sizeof(adapter->compare));
-		adapter->compare.source = &adapter->clock;
-		adapter->compare.target = ktime_get_real;
-		adapter->compare.num_samples = 10;
-		timecompare_update(&adapter->compare, 0);
-		break;
-	case e1000_82575:
-		/* 82575 does not support timesync */
-	default:
-		break;
-	}
-
 	dev_info(&pdev->dev, "Intel(R) Gigabit Ethernet Network Connection\n");
 	/* print bus type/speed/width info */
 	dev_info(&pdev->dev, "%s: (PCIe:%s:%s) %pM\n",
@@ -1781,6 +1731,70 @@ static void __devinit igb_probe_vfs(struct igb_adapter * adapter)
 #endif /* CONFIG_PCI_IOV */
 }
 
+
+/**
+ * igb_init_hw_timer - Initialize hardware timer used with IEEE 1588 timestamp
+ * @adapter: board private structure to initialize
+ *
+ * igb_init_hw_timer initializes the function pointer and values for the hw
+ * timer found in hardware.
+ **/
+static void igb_init_hw_timer(struct igb_adapter *adapter)
+{
+	struct e1000_hw *hw = &adapter->hw;
+
+	switch (hw->mac.type) {
+	case e1000_82576:
+		/*
+		 * Initialize hardware timer: we keep it running just in case
+		 * that some program needs it later on.
+		 */
+		memset(&adapter->cycles, 0, sizeof(adapter->cycles));
+		adapter->cycles.read = igb_read_clock;
+		adapter->cycles.mask = CLOCKSOURCE_MASK(64);
+		adapter->cycles.mult = 1;
+		/**
+		 * Scale the NIC clock cycle by a large factor so that
+		 * relatively small clock corrections can be added or
+		 * substracted at each clock tick. The drawbacks of a large
+		 * factor are a) that the clock register overflows more quickly
+		 * (not such a big deal) and b) that the increment per tick has
+		 * to fit into 24 bits.  As a result we need to use a shift of
+		 * 19 so we can fit a value of 16 into the TIMINCA register.
+		 */
+		adapter->cycles.shift = IGB_82576_TSYNC_SHIFT;
+		wr32(E1000_TIMINCA,
+		                (1 << E1000_TIMINCA_16NS_SHIFT) |
+		                (16 << IGB_82576_TSYNC_SHIFT));
+
+		/* Set registers so that rollover occurs soon to test this. */
+		wr32(E1000_SYSTIML, 0x00000000);
+		wr32(E1000_SYSTIMH, 0xFF800000);
+		wrfl();
+
+		timecounter_init(&adapter->clock,
+				 &adapter->cycles,
+				 ktime_to_ns(ktime_get_real()));
+		/*
+		 * Synchronize our NIC clock against system wall clock. NIC
+		 * time stamp reading requires ~3us per sample, each sample
+		 * was pretty stable even under load => only require 10
+		 * samples for each offset comparison.
+		 */
+		memset(&adapter->compare, 0, sizeof(adapter->compare));
+		adapter->compare.source = &adapter->clock;
+		adapter->compare.target = ktime_get_real;
+		adapter->compare.num_samples = 10;
+		timecompare_update(&adapter->compare, 0);
+		break;
+	case e1000_82575:
+		/* 82575 does not support timesync */
+	default:
+		break;
+	}
+
+}
+
 /**
  * igb_sw_init - Initialize general software structures (struct igb_adapter)
  * @adapter: board private structure to initialize
@@ -1816,6 +1830,7 @@ static int __devinit igb_sw_init(struct igb_adapter *adapter)
 		return -ENOMEM;
 	}
 
+	igb_init_hw_timer(adapter);
 	igb_probe_vfs(adapter);
 
 	/* Explicitly disable IRQ since the NIC can be in any state. */

[net-next-2.6 PATCH 3/7] igb: when number of CPUs > 4 combine tx/rx queues to allow more queues

[Jeff Kirsher]

From: Alexander Duyck <alexander.h.duyck@intel.com>

This patch makes it so that nics such as 82576 and newer can support more
hardware queues when there are more than 4 cpus by combining a tx/rx queue
pair onto one interrupt so that 8 queue pairs can be supported and thus
allow for more queues.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---

 drivers/net/igb/igb.h      |    9 +++++----
 drivers/net/igb/igb_main.c |   31 ++++++++++++++++++++++---------
 2 files changed, 27 insertions(+), 13 deletions(-)

diff --git a/drivers/net/igb/igb.h b/drivers/net/igb/igb.h
index 3298f5a..2bb9549 100644
--- a/drivers/net/igb/igb.h
+++ b/drivers/net/igb/igb.h
@@ -59,10 +59,10 @@ struct igb_adapter;
 #define MAX_Q_VECTORS                      8
 
 /* Transmit and receive queues */
-#define IGB_MAX_RX_QUEUES     (adapter->vfs_allocated_count ? \
-                               (adapter->vfs_allocated_count > 6 ? 1 : 2) : 4)
-#define IGB_MAX_TX_QUEUES     IGB_MAX_RX_QUEUES
-#define IGB_ABS_MAX_TX_QUEUES     4
+#define IGB_MAX_RX_QUEUES                  (adapter->vfs_allocated_count ? 2 : \
+                                           (hw->mac.type > e1000_82575 ? 8 : 4))
+#define IGB_ABS_MAX_TX_QUEUES              8
+#define IGB_MAX_TX_QUEUES                  IGB_MAX_RX_QUEUES
 
 #define IGB_MAX_VF_MC_ENTRIES              30
 #define IGB_MAX_VF_FUNCTIONS               8
@@ -315,6 +315,7 @@ struct igb_adapter {
 	u16 rx_ring_count;
 	unsigned int vfs_allocated_count;
 	struct vf_data_storage *vf_data;
+	u32 rss_queues;
 };
 
 #define IGB_FLAG_HAS_MSI           (1 << 0)
diff --git a/drivers/net/igb/igb_main.c b/drivers/net/igb/igb_main.c
index d72d484..0235220 100644
--- a/drivers/net/igb/igb_main.c
+++ b/drivers/net/igb/igb_main.c
@@ -296,10 +296,10 @@ static void igb_cache_ring_register(struct igb_adapter *adapter)
 		 * and continue consuming queues in the same sequence
 		 */
 		if (adapter->vfs_allocated_count) {
-			for (; i < adapter->num_rx_queues; i++)
+			for (; i < adapter->rss_queues; i++)
 				adapter->rx_ring[i].reg_idx = rbase_offset +
 				                              Q_IDX_82576(i);
-			for (; j < adapter->num_tx_queues; j++)
+			for (; j < adapter->rss_queues; j++)
 				adapter->tx_ring[j].reg_idx = rbase_offset +
 				                              Q_IDX_82576(j);
 		}
@@ -618,14 +618,15 @@ static void igb_set_interrupt_capability(struct igb_adapter *adapter)
 	int numvecs, i;
 
 	/* Number of supported queues. */
-	adapter->num_rx_queues = min_t(u32, IGB_MAX_RX_QUEUES, num_online_cpus());
-	adapter->num_tx_queues = min_t(u32, IGB_MAX_TX_QUEUES, num_online_cpus());
+	adapter->num_rx_queues = adapter->rss_queues;
+	adapter->num_tx_queues = adapter->rss_queues;
 
 	/* start with one vector for every rx queue */
 	numvecs = adapter->num_rx_queues;
 
 	/* if tx handler is seperate add 1 for every tx queue */
-	numvecs += adapter->num_tx_queues;
+	if (!(adapter->flags & IGB_FLAG_QUEUE_PAIRS))
+		numvecs += adapter->num_tx_queues;
 
 	/* store the number of vectors reserved for queues */
 	adapter->num_q_vectors = numvecs;
@@ -666,6 +667,7 @@ msi_only:
 	}
 #endif
 	adapter->vfs_allocated_count = 0;
+	adapter->rss_queues = 1;
 	adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
 	adapter->num_rx_queues = 1;
 	adapter->num_tx_queues = 1;
@@ -1824,6 +1826,17 @@ static int __devinit igb_sw_init(struct igb_adapter *adapter)
 		adapter->vfs_allocated_count = max_vfs;
 
 #endif /* CONFIG_PCI_IOV */
+	adapter->rss_queues = min_t(u32, IGB_MAX_RX_QUEUES, num_online_cpus());
+
+	/*
+	 * if rss_queues > 4 or vfs are going to be allocated with rss_queues
+	 * then we should combine the queues into a queue pair in order to
+	 * conserve interrupts due to limited supply
+	 */
+	if ((adapter->rss_queues > 4) ||
+	    ((adapter->rss_queues > 1) && (adapter->vfs_allocated_count > 6)))
+		adapter->flags |= IGB_FLAG_QUEUE_PAIRS;
+
 	/* This call may decrease the number of queues */
 	if (igb_init_interrupt_scheme(adapter)) {
 		dev_err(&pdev->dev, "Unable to allocate memory for queues\n");
@@ -2015,7 +2028,7 @@ static int igb_setup_all_tx_resources(struct igb_adapter *adapter)
 		}
 	}
 
-	for (i = 0; i < IGB_MAX_TX_QUEUES; i++) {
+	for (i = 0; i < IGB_ABS_MAX_TX_QUEUES; i++) {
 		int r_idx = i % adapter->num_tx_queues;
 		adapter->multi_tx_table[i] = &adapter->tx_ring[r_idx];
 	}
@@ -2199,7 +2212,7 @@ static void igb_setup_mrqc(struct igb_adapter *adapter)
 		array_wr32(E1000_RSSRK(0), j, rsskey);
 	}
 
-	num_rx_queues = adapter->num_rx_queues;
+	num_rx_queues = adapter->rss_queues;
 
 	if (adapter->vfs_allocated_count) {
 		/* 82575 and 82576 supports 2 RSS queues for VMDq */
@@ -2255,7 +2268,7 @@ static void igb_setup_mrqc(struct igb_adapter *adapter)
 				E1000_VT_CTL_DEFAULT_POOL_SHIFT;
 			wr32(E1000_VT_CTL, vtctl);
 		}
-		if (adapter->num_rx_queues > 1)
+		if (adapter->rss_queues > 1)
 			mrqc = E1000_MRQC_ENABLE_VMDQ_RSS_2Q;
 		else
 			mrqc = E1000_MRQC_ENABLE_VMDQ;
@@ -2385,7 +2398,7 @@ static inline void igb_set_vmolr(struct igb_adapter *adapter, int vfn)
 	/* clear all bits that might not be set */
 	vmolr &= ~(E1000_VMOLR_BAM | E1000_VMOLR_RSSE);
 
-	if (adapter->num_rx_queues > 1 && vfn == adapter->vfs_allocated_count)
+	if (adapter->rss_queues > 1 && vfn == adapter->vfs_allocated_count)
 		vmolr |= E1000_VMOLR_RSSE; /* enable RSS */
 	/*
 	 * for VMDq only allow the VFs and pool 0 to accept broadcast and

[net-next-2.6 PATCH 4/7] igb: Rework how netdev->stats is handled

[Jeff Kirsher]

From: Alexander Duyck <alexander.h.duyck@intel.com>

This patch does some refactoring work that I felt was needed after reviewing
the changes recently submitted relating to the replacement of net_stats with
netdev->stats.

This patch essentially creates two different collections of stats.  The
first handles the adapter specific states and is stored in gstring_stats,
and the second is for netdev specific stats and is stored in
gstring_net_stats.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---

 drivers/net/igb/igb_ethtool.c |  171 +++++++++++++++++++++--------------------
 drivers/net/igb/igb_main.c    |   40 +++++-----
 2 files changed, 108 insertions(+), 103 deletions(-)

diff --git a/drivers/net/igb/igb_ethtool.c b/drivers/net/igb/igb_ethtool.c
index f2b2825..c1cde5b 100644
--- a/drivers/net/igb/igb_ethtool.c
+++ b/drivers/net/igb/igb_ethtool.c
@@ -37,77 +37,88 @@
 
 #include "igb.h"
 
-enum {NETDEV_STATS, IGB_STATS};
-
 struct igb_stats {
 	char stat_string[ETH_GSTRING_LEN];
-	int type;
 	int sizeof_stat;
 	int stat_offset;
 };
 
-#define IGB_STAT(m)		IGB_STATS, \
-				FIELD_SIZEOF(struct igb_adapter, m), \
-				offsetof(struct igb_adapter, m)
-#define IGB_NETDEV_STAT(m)	NETDEV_STATS, \
-				FIELD_SIZEOF(struct net_device, m), \
-				offsetof(struct net_device, m)
-
+#define IGB_STAT(_name, _stat) { \
+	.stat_string = _name, \
+	.sizeof_stat = FIELD_SIZEOF(struct igb_adapter, _stat), \
+	.stat_offset = offsetof(struct igb_adapter, _stat) \
+}
 static const struct igb_stats igb_gstrings_stats[] = {
-	{ "rx_packets", IGB_STAT(stats.gprc) },
-	{ "tx_packets", IGB_STAT(stats.gptc) },
-	{ "rx_bytes", IGB_STAT(stats.gorc) },
-	{ "tx_bytes", IGB_STAT(stats.gotc) },
-	{ "rx_broadcast", IGB_STAT(stats.bprc) },
-	{ "tx_broadcast", IGB_STAT(stats.bptc) },
-	{ "rx_multicast", IGB_STAT(stats.mprc) },
-	{ "tx_multicast", IGB_STAT(stats.mptc) },
-	{ "rx_errors", IGB_NETDEV_STAT(stats.rx_errors) },
-	{ "tx_errors", IGB_NETDEV_STAT(stats.tx_errors) },
-	{ "tx_dropped", IGB_NETDEV_STAT(stats.tx_dropped) },
-	{ "multicast", IGB_STAT(stats.mprc) },
-	{ "collisions", IGB_STAT(stats.colc) },
-	{ "rx_length_errors", IGB_NETDEV_STAT(stats.rx_length_errors) },
-	{ "rx_over_errors", IGB_NETDEV_STAT(stats.rx_over_errors) },
-	{ "rx_crc_errors", IGB_STAT(stats.crcerrs) },
-	{ "rx_frame_errors", IGB_NETDEV_STAT(stats.rx_frame_errors) },
-	{ "rx_no_buffer_count", IGB_STAT(stats.rnbc) },
-	{ "rx_queue_drop_packet_count", IGB_NETDEV_STAT(stats.rx_fifo_errors) },
-	{ "rx_missed_errors", IGB_STAT(stats.mpc) },
-	{ "tx_aborted_errors", IGB_STAT(stats.ecol) },
-	{ "tx_carrier_errors", IGB_STAT(stats.tncrs) },
-	{ "tx_fifo_errors", IGB_NETDEV_STAT(stats.tx_fifo_errors) },
-	{ "tx_heartbeat_errors", IGB_NETDEV_STAT(stats.tx_heartbeat_errors) },
-	{ "tx_window_errors", IGB_STAT(stats.latecol) },
-	{ "tx_abort_late_coll", IGB_STAT(stats.latecol) },
-	{ "tx_deferred_ok", IGB_STAT(stats.dc) },
-	{ "tx_single_coll_ok", IGB_STAT(stats.scc) },
-	{ "tx_multi_coll_ok", IGB_STAT(stats.mcc) },
-	{ "tx_timeout_count", IGB_STAT(tx_timeout_count) },
-	{ "rx_long_length_errors", IGB_STAT(stats.roc) },
-	{ "rx_short_length_errors", IGB_STAT(stats.ruc) },
-	{ "rx_align_errors", IGB_STAT(stats.algnerrc) },
-	{ "tx_tcp_seg_good", IGB_STAT(stats.tsctc) },
-	{ "tx_tcp_seg_failed", IGB_STAT(stats.tsctfc) },
-	{ "rx_flow_control_xon", IGB_STAT(stats.xonrxc) },
-	{ "rx_flow_control_xoff", IGB_STAT(stats.xoffrxc) },
-	{ "tx_flow_control_xon", IGB_STAT(stats.xontxc) },
-	{ "tx_flow_control_xoff", IGB_STAT(stats.xofftxc) },
-	{ "rx_long_byte_count", IGB_STAT(stats.gorc) },
-	{ "tx_dma_out_of_sync", IGB_STAT(stats.doosync) },
-	{ "tx_smbus", IGB_STAT(stats.mgptc) },
-	{ "rx_smbus", IGB_STAT(stats.mgprc) },
-	{ "dropped_smbus", IGB_STAT(stats.mgpdc) },
+	IGB_STAT("rx_packets", stats.gprc),
+	IGB_STAT("tx_packets", stats.gptc),
+	IGB_STAT("rx_bytes", stats.gorc),
+	IGB_STAT("tx_bytes", stats.gotc),
+	IGB_STAT("rx_broadcast", stats.bprc),
+	IGB_STAT("tx_broadcast", stats.bptc),
+	IGB_STAT("rx_multicast", stats.mprc),
+	IGB_STAT("tx_multicast", stats.mptc),
+	IGB_STAT("multicast", stats.mprc),
+	IGB_STAT("collisions", stats.colc),
+	IGB_STAT("rx_crc_errors", stats.crcerrs),
+	IGB_STAT("rx_no_buffer_count", stats.rnbc),
+	IGB_STAT("rx_missed_errors", stats.mpc),
+	IGB_STAT("tx_aborted_errors", stats.ecol),
+	IGB_STAT("tx_carrier_errors", stats.tncrs),
+	IGB_STAT("tx_window_errors", stats.latecol),
+	IGB_STAT("tx_abort_late_coll", stats.latecol),
+	IGB_STAT("tx_deferred_ok", stats.dc),
+	IGB_STAT("tx_single_coll_ok", stats.scc),
+	IGB_STAT("tx_multi_coll_ok", stats.mcc),
+	IGB_STAT("tx_timeout_count", tx_timeout_count),
+	IGB_STAT("rx_long_length_errors", stats.roc),
+	IGB_STAT("rx_short_length_errors", stats.ruc),
+	IGB_STAT("rx_align_errors", stats.algnerrc),
+	IGB_STAT("tx_tcp_seg_good", stats.tsctc),
+	IGB_STAT("tx_tcp_seg_failed", stats.tsctfc),
+	IGB_STAT("rx_flow_control_xon", stats.xonrxc),
+	IGB_STAT("rx_flow_control_xoff", stats.xoffrxc),
+	IGB_STAT("tx_flow_control_xon", stats.xontxc),
+	IGB_STAT("tx_flow_control_xoff", stats.xofftxc),
+	IGB_STAT("rx_long_byte_count", stats.gorc),
+	IGB_STAT("tx_dma_out_of_sync", stats.doosync),
+	IGB_STAT("tx_smbus", stats.mgptc),
+	IGB_STAT("rx_smbus", stats.mgprc),
+	IGB_STAT("dropped_smbus", stats.mgpdc),
+};
+
+#define IGB_NETDEV_STAT(_net_stat) { \
+	.stat_string = __stringify(_net_stat), \
+	.sizeof_stat = FIELD_SIZEOF(struct net_device_stats, _net_stat), \
+	.stat_offset = offsetof(struct net_device_stats, _net_stat) \
+}
+static const struct igb_stats igb_gstrings_net_stats[] = {
+	IGB_NETDEV_STAT(rx_errors),
+	IGB_NETDEV_STAT(tx_errors),
+	IGB_NETDEV_STAT(tx_dropped),
+	IGB_NETDEV_STAT(rx_length_errors),
+	IGB_NETDEV_STAT(rx_over_errors),
+	IGB_NETDEV_STAT(rx_frame_errors),
+	IGB_NETDEV_STAT(rx_fifo_errors),
+	IGB_NETDEV_STAT(tx_fifo_errors),
+	IGB_NETDEV_STAT(tx_heartbeat_errors)
 };
 
+#define IGB_GLOBAL_STATS_LEN	\
+	(sizeof(igb_gstrings_stats) / sizeof(struct igb_stats))
+#define IGB_NETDEV_STATS_LEN	\
+	(sizeof(igb_gstrings_net_stats) / sizeof(struct igb_stats))
+#define IGB_RX_QUEUE_STATS_LEN \
+	(sizeof(struct igb_rx_queue_stats) / sizeof(u64))
+#define IGB_TX_QUEUE_STATS_LEN \
+	(sizeof(struct igb_tx_queue_stats) / sizeof(u64))
 #define IGB_QUEUE_STATS_LEN \
 	((((struct igb_adapter *)netdev_priv(netdev))->num_rx_queues * \
-	  (sizeof(struct igb_rx_queue_stats) / sizeof(u64))) + \
+	  IGB_RX_QUEUE_STATS_LEN) + \
 	 (((struct igb_adapter *)netdev_priv(netdev))->num_tx_queues * \
-	  (sizeof(struct igb_tx_queue_stats) / sizeof(u64))))
-#define IGB_GLOBAL_STATS_LEN	\
-	(sizeof(igb_gstrings_stats) / sizeof(struct igb_stats))
-#define IGB_STATS_LEN (IGB_GLOBAL_STATS_LEN + IGB_QUEUE_STATS_LEN)
+	  IGB_TX_QUEUE_STATS_LEN))
+#define IGB_STATS_LEN \
+	(IGB_GLOBAL_STATS_LEN + IGB_NETDEV_STATS_LEN + IGB_QUEUE_STATS_LEN)
+
 static const char igb_gstrings_test[][ETH_GSTRING_LEN] = {
 	"Register test  (offline)", "Eeprom test    (offline)",
 	"Interrupt test (offline)", "Loopback test  (offline)",
@@ -1922,43 +1933,32 @@ static void igb_get_ethtool_stats(struct net_device *netdev,
 				  struct ethtool_stats *stats, u64 *data)
 {
 	struct igb_adapter *adapter = netdev_priv(netdev);
+	struct net_device_stats *net_stats = &netdev->stats;
 	u64 *queue_stat;
-	int stat_count_tx = sizeof(struct igb_tx_queue_stats) / sizeof(u64);
-	int stat_count_rx = sizeof(struct igb_rx_queue_stats) / sizeof(u64);
-	int j;
-	int i;
-	char *p = NULL;
+	int i, j, k;
+	char *p;
 
 	igb_update_stats(adapter);
 
 	for (i = 0; i < IGB_GLOBAL_STATS_LEN; i++) {
-		switch (igb_gstrings_stats[i].type) {
-		case NETDEV_STATS:
-			p = (char *) netdev +
-					igb_gstrings_stats[i].stat_offset;
-			break;
-		case IGB_STATS:
-			p = (char *) adapter +
-					igb_gstrings_stats[i].stat_offset;
-			break;
-		}
-
+		p = (char *)adapter + igb_gstrings_stats[i].stat_offset;
 		data[i] = (igb_gstrings_stats[i].sizeof_stat ==
 			sizeof(u64)) ? *(u64 *)p : *(u32 *)p;
 	}
+	for (j = 0; j < IGB_NETDEV_STATS_LEN; j++, i++) {
+		p = (char *)net_stats + igb_gstrings_net_stats[j].stat_offset;
+		data[i] = (igb_gstrings_net_stats[j].sizeof_stat ==
+			sizeof(u64)) ? *(u64 *)p : *(u32 *)p;
+	}
 	for (j = 0; j < adapter->num_tx_queues; j++) {
-		int k;
 		queue_stat = (u64 *)&adapter->tx_ring[j].tx_stats;
-		for (k = 0; k < stat_count_tx; k++)
-			data[i + k] = queue_stat[k];
-		i += k;
+		for (k = 0; k < IGB_TX_QUEUE_STATS_LEN; k++, i++)
+			data[i] = queue_stat[k];
 	}
 	for (j = 0; j < adapter->num_rx_queues; j++) {
-		int k;
 		queue_stat = (u64 *)&adapter->rx_ring[j].rx_stats;
-		for (k = 0; k < stat_count_rx; k++)
-			data[i + k] = queue_stat[k];
-		i += k;
+		for (k = 0; k < IGB_RX_QUEUE_STATS_LEN; k++, i++)
+			data[i] = queue_stat[k];
 	}
 }
 
@@ -1979,6 +1979,11 @@ static void igb_get_strings(struct net_device *netdev, u32 stringset, u8 *data)
 			       ETH_GSTRING_LEN);
 			p += ETH_GSTRING_LEN;
 		}
+		for (i = 0; i < IGB_NETDEV_STATS_LEN; i++) {
+			memcpy(p, igb_gstrings_net_stats[i].stat_string,
+			       ETH_GSTRING_LEN);
+			p += ETH_GSTRING_LEN;
+		}
 		for (i = 0; i < adapter->num_tx_queues; i++) {
 			sprintf(p, "tx_queue_%u_packets", i);
 			p += ETH_GSTRING_LEN;
diff --git a/drivers/net/igb/igb_main.c b/drivers/net/igb/igb_main.c
index 0235220..4d4ab87 100644
--- a/drivers/net/igb/igb_main.c
+++ b/drivers/net/igb/igb_main.c
@@ -3789,7 +3789,7 @@ static int igb_change_mtu(struct net_device *netdev, int new_mtu)
 
 void igb_update_stats(struct igb_adapter *adapter)
 {
-	struct net_device *netdev = adapter->netdev;
+	struct net_device_stats *net_stats = igb_get_stats(adapter->netdev);
 	struct e1000_hw *hw = &adapter->hw;
 	struct pci_dev *pdev = adapter->pdev;
 	u32 rnbc;
@@ -3813,13 +3813,13 @@ void igb_update_stats(struct igb_adapter *adapter)
 	for (i = 0; i < adapter->num_rx_queues; i++) {
 		u32 rqdpc_tmp = rd32(E1000_RQDPC(i)) & 0x0FFF;
 		adapter->rx_ring[i].rx_stats.drops += rqdpc_tmp;
-		netdev->stats.rx_fifo_errors += rqdpc_tmp;
+		net_stats->rx_fifo_errors += rqdpc_tmp;
 		bytes += adapter->rx_ring[i].rx_stats.bytes;
 		packets += adapter->rx_ring[i].rx_stats.packets;
 	}
 
-	netdev->stats.rx_bytes = bytes;
-	netdev->stats.rx_packets = packets;
+	net_stats->rx_bytes = bytes;
+	net_stats->rx_packets = packets;
 
 	bytes = 0;
 	packets = 0;
@@ -3827,8 +3827,8 @@ void igb_update_stats(struct igb_adapter *adapter)
 		bytes += adapter->tx_ring[i].tx_stats.bytes;
 		packets += adapter->tx_ring[i].tx_stats.packets;
 	}
-	netdev->stats.tx_bytes = bytes;
-	netdev->stats.tx_packets = packets;
+	net_stats->tx_bytes = bytes;
+	net_stats->tx_packets = packets;
 
 	/* read stats registers */
 	adapter->stats.crcerrs += rd32(E1000_CRCERRS);
@@ -3865,7 +3865,7 @@ void igb_update_stats(struct igb_adapter *adapter)
 	rd32(E1000_GOTCH); /* clear GOTCL */
 	rnbc = rd32(E1000_RNBC);
 	adapter->stats.rnbc += rnbc;
-	netdev->stats.rx_fifo_errors += rnbc;
+	net_stats->rx_fifo_errors += rnbc;
 	adapter->stats.ruc += rd32(E1000_RUC);
 	adapter->stats.rfc += rd32(E1000_RFC);
 	adapter->stats.rjc += rd32(E1000_RJC);
@@ -3906,29 +3906,29 @@ void igb_update_stats(struct igb_adapter *adapter)
 	adapter->stats.icrxdmtc += rd32(E1000_ICRXDMTC);
 
 	/* Fill out the OS statistics structure */
-	netdev->stats.multicast = adapter->stats.mprc;
-	netdev->stats.collisions = adapter->stats.colc;
+	net_stats->multicast = adapter->stats.mprc;
+	net_stats->collisions = adapter->stats.colc;
 
 	/* Rx Errors */
 
 	/* RLEC on some newer hardware can be incorrect so build
 	 * our own version based on RUC and ROC */
-	netdev->stats.rx_errors = adapter->stats.rxerrc +
+	net_stats->rx_errors = adapter->stats.rxerrc +
 		adapter->stats.crcerrs + adapter->stats.algnerrc +
 		adapter->stats.ruc + adapter->stats.roc +
 		adapter->stats.cexterr;
-	netdev->stats.rx_length_errors = adapter->stats.ruc +
-					      adapter->stats.roc;
-	netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
-	netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
-	netdev->stats.rx_missed_errors = adapter->stats.mpc;
+	net_stats->rx_length_errors = adapter->stats.ruc +
+				      adapter->stats.roc;
+	net_stats->rx_crc_errors = adapter->stats.crcerrs;
+	net_stats->rx_frame_errors = adapter->stats.algnerrc;
+	net_stats->rx_missed_errors = adapter->stats.mpc;
 
 	/* Tx Errors */
-	netdev->stats.tx_errors = adapter->stats.ecol +
-				       adapter->stats.latecol;
-	netdev->stats.tx_aborted_errors = adapter->stats.ecol;
-	netdev->stats.tx_window_errors = adapter->stats.latecol;
-	netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
+	net_stats->tx_errors = adapter->stats.ecol +
+			       adapter->stats.latecol;
+	net_stats->tx_aborted_errors = adapter->stats.ecol;
+	net_stats->tx_window_errors = adapter->stats.latecol;
+	net_stats->tx_carrier_errors = adapter->stats.tncrs;
 
 	/* Tx Dropped needs to be maintained elsewhere */
 

[net-next-2.6 PATCH 5/7] igb: removed unused tx/rx total bytes/packets from adapter struct

[Jeff Kirsher]

From: Alexander Duyck <alexander.h.duyck@intel.com>

This patch removes unused variables total_tx_bytes, total_tx_packets,
total_rx_bytes, and total_rx_packets from the adapter struct.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---

 drivers/net/igb/igb.h |    4 ----
 1 files changed, 0 insertions(+), 4 deletions(-)

diff --git a/drivers/net/igb/igb.h b/drivers/net/igb/igb.h
index 2bb9549..63abd1c 100644
--- a/drivers/net/igb/igb.h
+++ b/drivers/net/igb/igb.h
@@ -249,10 +249,6 @@ struct igb_adapter {
 	u16 link_speed;
 	u16 link_duplex;
 
-	unsigned int total_tx_bytes;
-	unsigned int total_tx_packets;
-	unsigned int total_rx_bytes;
-	unsigned int total_rx_packets;
 	/* Interrupt Throttle Rate */
 	u32 rx_itr_setting;
 	u32 tx_itr_setting;

[net-next-2.6 PATCH 6/7] igb: check for packets on all tx rings when link is down

[Jeff Kirsher]

From: Alexander Duyck <alexander.h.duyck@intel.com>

We were previously only checking the first tx ring to see if it had any
packets in it when the link when down.  However we should be checking all
of the rings so this patch makes it so that all of the rings are now being
checked.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---

 drivers/net/igb/igb_main.c |   23 ++++++++++++-----------
 1 files changed, 12 insertions(+), 11 deletions(-)

diff --git a/drivers/net/igb/igb_main.c b/drivers/net/igb/igb_main.c
index 4d4ab87..9911b3a 100644
--- a/drivers/net/igb/igb_main.c
+++ b/drivers/net/igb/igb_main.c
@@ -2943,7 +2943,6 @@ static void igb_watchdog_task(struct work_struct *work)
                                                    watchdog_task);
 	struct e1000_hw *hw = &adapter->hw;
 	struct net_device *netdev = adapter->netdev;
-	struct igb_ring *tx_ring = adapter->tx_ring;
 	u32 link;
 	int i;
 
@@ -3013,22 +3012,24 @@ static void igb_watchdog_task(struct work_struct *work)
 	igb_update_stats(adapter);
 	igb_update_adaptive(hw);
 
-	if (!netif_carrier_ok(netdev)) {
-		if (igb_desc_unused(tx_ring) + 1 < tx_ring->count) {
+	for (i = 0; i < adapter->num_tx_queues; i++) {
+		struct igb_ring *tx_ring = &adapter->tx_ring[i];
+		if (!netif_carrier_ok(netdev)) {
 			/* We've lost link, so the controller stops DMA,
 			 * but we've got queued Tx work that's never going
 			 * to get done, so reset controller to flush Tx.
 			 * (Do the reset outside of interrupt context). */
-			adapter->tx_timeout_count++;
-			schedule_work(&adapter->reset_task);
-			/* return immediately since reset is imminent */
-			return;
+			if (igb_desc_unused(tx_ring) + 1 < tx_ring->count) {
+				adapter->tx_timeout_count++;
+				schedule_work(&adapter->reset_task);
+				/* return immediately since reset is imminent */
+				return;
+			}
 		}
-	}
 
-	/* Force detection of hung controller every watchdog period */
-	for (i = 0; i < adapter->num_tx_queues; i++)
-		adapter->tx_ring[i].detect_tx_hung = true;
+		/* Force detection of hung controller every watchdog period */
+		tx_ring->detect_tx_hung = true;
+	}
 
 	/* Cause software interrupt to ensure rx ring is cleaned */
 	if (adapter->msix_entries) {

[net-next-2.6 PATCH 7/7] igb: only recycle page if it is on our numa node

[Jeff Kirsher]

From: Alexander Duyck <alexander.h.duyck@intel.com>

This patch makes it so that we only recycle pages when they are from the
local NUMA node.  Non-local pages are freed and replaced with locally
allocated pages.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---

 drivers/net/igb/igb_main.c |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)

diff --git a/drivers/net/igb/igb_main.c b/drivers/net/igb/igb_main.c
index 9911b3a..0cab5e2 100644
--- a/drivers/net/igb/igb_main.c
+++ b/drivers/net/igb/igb_main.c
@@ -4952,6 +4952,7 @@ static bool igb_clean_rx_irq_adv(struct igb_q_vector *q_vector,
 	struct sk_buff *skb;
 	bool cleaned = false;
 	int cleaned_count = 0;
+	int current_node = numa_node_id();
 	unsigned int total_bytes = 0, total_packets = 0;
 	unsigned int i;
 	u32 staterr;
@@ -5006,7 +5007,8 @@ static bool igb_clean_rx_irq_adv(struct igb_q_vector *q_vector,
 						buffer_info->page_offset,
 						length);
 
-			if (page_count(buffer_info->page) != 1)
+			if ((page_count(buffer_info->page) != 1) ||
+			    (page_to_nid(buffer_info->page) != current_node))
 				buffer_info->page = NULL;
 			else
 				get_page(buffer_info->page);

Re: [net-next-2.6 PATCH 1/7] igb: change type for ring sizes to u16 in igb_set_ring_param

[David Miller]

All 7 patches applied, thanks.