1 files changed, 138 insertions, 155 deletions

diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
index 7e91d825c760..47e9a90d6b10 100644
--- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
@@ -140,65 +140,6 @@ static inline u32 i40e_get_head(struct i40e_ring *tx_ring)
 	return le32_to_cpu(*(volatile __le32 *)head);
 }
 
-/**
- * i40e_get_tx_pending - how many tx descriptors not processed
- * @tx_ring: the ring of descriptors
- *
- * Since there is no access to the ring head register
- * in XL710, we need to use our local copies
- **/
-static u32 i40e_get_tx_pending(struct i40e_ring *ring)
-{
-	u32 head, tail;
-
-	head = i40e_get_head(ring);
-	tail = readl(ring->tail);
-
-	if (head != tail)
-		return (head < tail) ?
-			tail - head : (tail + ring->count - head);
-
-	return 0;
-}
-
-/**
- * i40e_check_tx_hang - Is there a hang in the Tx queue
- * @tx_ring: the ring of descriptors
- **/
-static bool i40e_check_tx_hang(struct i40e_ring *tx_ring)
-{
-	u32 tx_done = tx_ring->stats.packets;
-	u32 tx_done_old = tx_ring->tx_stats.tx_done_old;
-	u32 tx_pending = i40e_get_tx_pending(tx_ring);
-	bool ret = false;
-
-	clear_check_for_tx_hang(tx_ring);
-
-	/* Check for a hung queue, but be thorough. This verifies
-	 * that a transmit has been completed since the previous
-	 * check AND there is at least one packet pending. The
-	 * ARMED bit is set to indicate a potential hang. The
-	 * bit is cleared if a pause frame is received to remove
-	 * false hang detection due to PFC or 802.3x frames. By
-	 * requiring this to fail twice we avoid races with
-	 * PFC clearing the ARMED bit and conditions where we
-	 * run the check_tx_hang logic with a transmit completion
-	 * pending but without time to complete it yet.
-	 */
-	if ((tx_done_old == tx_done) && tx_pending) {
-		/* make sure it is true for two checks in a row */
-		ret = test_and_set_bit(__I40E_HANG_CHECK_ARMED,
-				       &tx_ring->state);
-	} else if (tx_done_old == tx_done &&
-		   (tx_pending < I40E_MIN_DESC_PENDING) && (tx_pending > 0)) {
-		/* update completed stats and disarm the hang check */
-		tx_ring->tx_stats.tx_done_old = tx_done;
-		clear_bit(__I40E_HANG_CHECK_ARMED, &tx_ring->state);
-	}
-
-	return ret;
-}
-
 #define WB_STRIDE 0x3
 
 /**
@@ -304,36 +245,15 @@ static bool i40e_clean_tx_irq(struct i40e_ring *tx_ring, int budget)
 	tx_ring->q_vector->tx.total_bytes += total_bytes;
 	tx_ring->q_vector->tx.total_packets += total_packets;
 
+	/* check to see if there are any non-cache aligned descriptors
+	 * waiting to be written back, and kick the hardware to force
+	 * them to be written back in case of napi polling
+	 */
 	if (budget &&
 	    !((i & WB_STRIDE) == WB_STRIDE) &&
 	    !test_bit(__I40E_DOWN, &tx_ring->vsi->state) &&
 	    (I40E_DESC_UNUSED(tx_ring) != tx_ring->count))
 		tx_ring->arm_wb = true;
-	else
-		tx_ring->arm_wb = false;
-
-	if (check_for_tx_hang(tx_ring) && i40e_check_tx_hang(tx_ring)) {
-		/* schedule immediate reset if we believe we hung */
-		dev_info(tx_ring->dev, "Detected Tx Unit Hang\n"
-			 "  VSI                  <%d>\n"
-			 "  Tx Queue             <%d>\n"
-			 "  next_to_use          <%x>\n"
-			 "  next_to_clean        <%x>\n",
-			 tx_ring->vsi->seid,
-			 tx_ring->queue_index,
-			 tx_ring->next_to_use, i);
-
-		netif_stop_subqueue(tx_ring->netdev, tx_ring->queue_index);
-
-		dev_info(tx_ring->dev,
-			 "tx hang detected on queue %d, resetting adapter\n",
-			 tx_ring->queue_index);
-
-		tx_ring->netdev->netdev_ops->ndo_tx_timeout(tx_ring->netdev);
-
-		/* the adapter is about to reset, no point in enabling stuff */
-		return true;
-	}
 
 	netdev_tx_completed_queue(netdev_get_tx_queue(tx_ring->netdev,
 						      tx_ring->queue_index),
@@ -355,16 +275,16 @@ static bool i40e_clean_tx_irq(struct i40e_ring *tx_ring, int budget)
 		}
 	}
 
-	return budget > 0;
+	return !!budget;
 }
 
 /**
- * i40e_force_wb -Arm hardware to do a wb on noncache aligned descriptors
+ * i40evf_force_wb -Arm hardware to do a wb on noncache aligned descriptors
  * @vsi: the VSI we care about
  * @q_vector: the vector  on which to force writeback
  *
  **/
-static void i40e_force_wb(struct i40e_vsi *vsi, struct i40e_q_vector *q_vector)
+static void i40evf_force_wb(struct i40e_vsi *vsi, struct i40e_q_vector *q_vector)
 {
 	u16 flags = q_vector->tx.ring[0].flags;
 
@@ -398,6 +318,8 @@ static void i40e_force_wb(struct i40e_vsi *vsi, struct i40e_q_vector *q_vector)
  * i40e_set_new_dynamic_itr - Find new ITR level
  * @rc: structure containing ring performance data
  *
+ * Returns true if ITR changed, false if not
+ *
  * Stores a new ITR value based on packets and byte counts during
  * the last interrupt.  The advantage of per interrupt computation
  * is faster updates and more accurate ITR for the current traffic
@@ -406,21 +328,32 @@ static void i40e_force_wb(struct i40e_vsi *vsi, struct i40e_q_vector *q_vector)
  * testing data as well as attempting to minimize response time
  * while increasing bulk throughput.
  **/
-static void i40e_set_new_dynamic_itr(struct i40e_ring_container *rc)
+static bool i40e_set_new_dynamic_itr(struct i40e_ring_container *rc)
 {
 	enum i40e_latency_range new_latency_range = rc->latency_range;
+	struct i40e_q_vector *qv = rc->ring->q_vector;
 	u32 new_itr = rc->itr;
 	int bytes_per_int;
+	int usecs;
 
 	if (rc->total_packets == 0 || !rc->itr)
-		return;
+		return false;
 
 	/* simple throttlerate management
-	 *   0-10MB/s   lowest (100000 ints/s)
+	 *   0-10MB/s   lowest (50000 ints/s)
 	 *  10-20MB/s   low    (20000 ints/s)
-	 *  20-1249MB/s bulk   (8000 ints/s)
+	 *  20-1249MB/s bulk   (18000 ints/s)
+	 *  > 40000 Rx packets per second (8000 ints/s)
+	 *
+	 * The math works out because the divisor is in 10^(-6) which
+	 * turns the bytes/us input value into MB/s values, but
+	 * make sure to use usecs, as the register values written
+	 * are in 2 usec increments in the ITR registers, and make sure
+	 * to use the smoothed values that the countdown timer gives us.
 	 */
-	bytes_per_int = rc->total_bytes / rc->itr;
+	usecs = (rc->itr << 1) * ITR_COUNTDOWN_START;
+	bytes_per_int = rc->total_bytes / usecs;
+
 	switch (new_latency_range) {
 	case I40E_LOWEST_LATENCY:
 		if (bytes_per_int > 10)
@@ -433,35 +366,52 @@ static void i40e_set_new_dynamic_itr(struct i40e_ring_container *rc)
 			new_latency_range = I40E_LOWEST_LATENCY;
 		break;
 	case I40E_BULK_LATENCY:
-		if (bytes_per_int <= 20)
-			new_latency_range = I40E_LOW_LATENCY;
-		break;
+	case I40E_ULTRA_LATENCY:
 	default:
 		if (bytes_per_int <= 20)
 			new_latency_range = I40E_LOW_LATENCY;
 		break;
 	}
+
+	/* this is to adjust RX more aggressively when streaming small
+	 * packets.  The value of 40000 was picked as it is just beyond
+	 * what the hardware can receive per second if in low latency
+	 * mode.
+	 */
+#define RX_ULTRA_PACKET_RATE 40000
+
+	if ((((rc->total_packets * 1000000) / usecs) > RX_ULTRA_PACKET_RATE) &&
+	    (&qv->rx == rc))
+		new_latency_range = I40E_ULTRA_LATENCY;
+
 	rc->latency_range = new_latency_range;
 
 	switch (new_latency_range) {
 	case I40E_LOWEST_LATENCY:
-		new_itr = I40E_ITR_100K;
+		new_itr = I40E_ITR_50K;
 		break;
 	case I40E_LOW_LATENCY:
 		new_itr = I40E_ITR_20K;
 		break;
 	case I40E_BULK_LATENCY:
+		new_itr = I40E_ITR_18K;
+		break;
+	case I40E_ULTRA_LATENCY:
 		new_itr = I40E_ITR_8K;
 		break;
 	default:
 		break;
 	}
 
-	if (new_itr != rc->itr)
-		rc->itr = new_itr;
-
 	rc->total_bytes = 0;
 	rc->total_packets = 0;
+
+	if (new_itr != rc->itr) {
+		rc->itr = new_itr;
+		return true;
+	}
+
+	return false;
 }
 
 /*
@@ -822,16 +772,11 @@ static void i40e_receive_skb(struct i40e_ring *rx_ring,
 			     struct sk_buff *skb, u16 vlan_tag)
 {
 	struct i40e_q_vector *q_vector = rx_ring->q_vector;
-	struct i40e_vsi *vsi = rx_ring->vsi;
-	u64 flags = vsi->back->flags;
 
 	if (vlan_tag & VLAN_VID_MASK)
 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), vlan_tag);
 
-	if (flags & I40E_FLAG_IN_NETPOLL)
-		netif_rx(skb);
-	else
-		napi_gro_receive(&q_vector->napi, skb);
+	napi_gro_receive(&q_vector->napi, skb);
 }
 
 /**
@@ -997,7 +942,7 @@ static int i40e_clean_rx_irq_ps(struct i40e_ring *rx_ring, int budget)
 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
 	u16 rx_packet_len, rx_header_len, rx_sph, rx_hbo;
 	u16 cleaned_count = I40E_DESC_UNUSED(rx_ring);
-	const int current_node = numa_node_id();
+	const int current_node = numa_mem_id();
 	struct i40e_vsi *vsi = rx_ring->vsi;
 	u16 i = rx_ring->next_to_clean;
 	union i40e_rx_desc *rx_desc;
@@ -1067,6 +1012,7 @@ static int i40e_clean_rx_irq_ps(struct i40e_ring *rx_ring, int budget)
 		cleaned_count++;
 		if (rx_hbo || rx_sph) {
 			int len;
+
 			if (rx_hbo)
 				len = I40E_RX_HDR_SIZE;
 			else
@@ -1240,9 +1186,6 @@ static int i40e_clean_rx_irq_1buf(struct i40e_ring *rx_ring, int budget)
 		/* ERR_MASK will only have valid bits if EOP set */
 		if (unlikely(rx_error & BIT(I40E_RX_DESC_ERROR_RXE_SHIFT))) {
 			dev_kfree_skb_any(skb);
-			/* TODO: shouldn't we increment a counter indicating the
-			 * drop?
-			 */
 			continue;
 		}
 
@@ -1274,6 +1217,21 @@ static int i40e_clean_rx_irq_1buf(struct i40e_ring *rx_ring, int budget)
 	return total_rx_packets;
 }
 
+static u32 i40e_buildreg_itr(const int type, const u16 itr)
+{
+	u32 val;
+
+	val = I40E_VFINT_DYN_CTLN1_INTENA_MASK |
+	      I40E_VFINT_DYN_CTLN1_CLEARPBA_MASK |
+	      (type << I40E_VFINT_DYN_CTLN1_ITR_INDX_SHIFT) |
+	      (itr << I40E_VFINT_DYN_CTLN1_INTERVAL_SHIFT);
+
+	return val;
+}
+
+/* a small macro to shorten up some long lines */
+#define INTREG I40E_VFINT_DYN_CTLN1
+
 /**
  * i40e_update_enable_itr - Update itr and re-enable MSIX interrupt
  * @vsi: the VSI we care about
@@ -1284,55 +1242,67 @@ static inline void i40e_update_enable_itr(struct i40e_vsi *vsi,
 					  struct i40e_q_vector *q_vector)
 {
 	struct i40e_hw *hw = &vsi->back->hw;
-	u16 old_itr;
+	bool rx = false, tx = false;
+	u32 rxval, txval;
 	int vector;
-	u32 val;
 
 	vector = (q_vector->v_idx + vsi->base_vector);
+
+	/* avoid dynamic calculation if in countdown mode OR if
+	 * all dynamic is disabled
+	 */
+	rxval = txval = i40e_buildreg_itr(I40E_ITR_NONE, 0);
+
+	if (q_vector->itr_countdown > 0 ||
+	    (!ITR_IS_DYNAMIC(vsi->rx_itr_setting) &&
+	     !ITR_IS_DYNAMIC(vsi->tx_itr_setting))) {
+		goto enable_int;
+	}
+
 	if (ITR_IS_DYNAMIC(vsi->rx_itr_setting)) {
-		old_itr = q_vector->rx.itr;
-		i40e_set_new_dynamic_itr(&q_vector->rx);
-		if (old_itr != q_vector->rx.itr) {
-			val = I40E_VFINT_DYN_CTLN1_INTENA_MASK |
-			I40E_VFINT_DYN_CTLN1_CLEARPBA_MASK |
-			(I40E_RX_ITR <<
-				I40E_VFINT_DYN_CTLN1_ITR_INDX_SHIFT) |
-			(q_vector->rx.itr <<
-				I40E_VFINT_DYN_CTLN1_INTERVAL_SHIFT);
-		} else {
-			val = I40E_VFINT_DYN_CTLN1_INTENA_MASK |
-			I40E_VFINT_DYN_CTLN1_CLEARPBA_MASK |
-			(I40E_ITR_NONE <<
-				I40E_VFINT_DYN_CTLN1_ITR_INDX_SHIFT);
-		}
-		if (!test_bit(__I40E_DOWN, &vsi->state))
-			wr32(hw, I40E_VFINT_DYN_CTLN1(vector - 1), val);
-	} else {
-		i40evf_irq_enable_queues(vsi->back, 1
-			<< q_vector->v_idx);
+		rx = i40e_set_new_dynamic_itr(&q_vector->rx);
+		rxval = i40e_buildreg_itr(I40E_RX_ITR, q_vector->rx.itr);
 	}
 	if (ITR_IS_DYNAMIC(vsi->tx_itr_setting)) {
-		old_itr = q_vector->tx.itr;
-		i40e_set_new_dynamic_itr(&q_vector->tx);
-		if (old_itr != q_vector->tx.itr) {
-			val = I40E_VFINT_DYN_CTLN1_INTENA_MASK |
-				I40E_VFINT_DYN_CTLN1_CLEARPBA_MASK |
-				(I40E_TX_ITR <<
-				   I40E_VFINT_DYN_CTLN1_ITR_INDX_SHIFT) |
-				(q_vector->tx.itr <<
-				   I40E_VFINT_DYN_CTLN1_INTERVAL_SHIFT);
+		tx = i40e_set_new_dynamic_itr(&q_vector->tx);
+		txval = i40e_buildreg_itr(I40E_TX_ITR, q_vector->tx.itr);
+	}
+	if (rx || tx) {
+		/* get the higher of the two ITR adjustments and
+		 * use the same value for both ITR registers
+		 * when in adaptive mode (Rx and/or Tx)
+		 */
+		u16 itr = max(q_vector->tx.itr, q_vector->rx.itr);
 
-		} else {
-			val = I40E_VFINT_DYN_CTLN1_INTENA_MASK |
-				I40E_VFINT_DYN_CTLN1_CLEARPBA_MASK |
-				(I40E_ITR_NONE <<
-				   I40E_VFINT_DYN_CTLN1_ITR_INDX_SHIFT);
-		}
-		if (!test_bit(__I40E_DOWN, &vsi->state))
-			wr32(hw, I40E_VFINT_DYN_CTLN1(vector - 1), val);
-	} else {
-		i40evf_irq_enable_queues(vsi->back, BIT(q_vector->v_idx));
+		q_vector->tx.itr = q_vector->rx.itr = itr;
+		txval = i40e_buildreg_itr(I40E_TX_ITR, itr);
+		tx = true;
+		rxval = i40e_buildreg_itr(I40E_RX_ITR, itr);
+		rx = true;
 	}
+
+	/* only need to enable the interrupt once, but need
+	 * to possibly update both ITR values
+	 */
+	if (rx) {
+		/* set the INTENA_MSK_MASK so that this first write
+		 * won't actually enable the interrupt, instead just
+		 * updating the ITR (it's bit 31 PF and VF)
+		 */
+		rxval |= BIT(31);
+		/* don't check _DOWN because interrupt isn't being enabled */
+		wr32(hw, INTREG(vector - 1), rxval);
+	}
+
+enable_int:
+	if (!test_bit(__I40E_DOWN, &vsi->state))
+		wr32(hw, INTREG(vector - 1), txval);
+
+	if (q_vector->itr_countdown)
+		q_vector->itr_countdown--;
+	else
+		q_vector->itr_countdown = ITR_COUNTDOWN_START;
+
 }
 
 /**
@@ -1353,7 +1323,7 @@ int i40evf_napi_poll(struct napi_struct *napi, int budget)
 	bool clean_complete = true;
 	bool arm_wb = false;
 	int budget_per_ring;
-	int cleaned;
+	int work_done = 0;
 
 	if (test_bit(__I40E_DOWN, &vsi->state)) {
 		napi_complete(napi);
@@ -1366,26 +1336,36 @@ int i40evf_napi_poll(struct napi_struct *napi, int budget)
 	i40e_for_each_ring(ring, q_vector->tx) {
 		clean_complete &= i40e_clean_tx_irq(ring, vsi->work_limit);
 		arm_wb |= ring->arm_wb;
+		ring->arm_wb = false;
 	}
 
+	/* Handle case where we are called by netpoll with a budget of 0 */
+	if (budget <= 0)
+		goto tx_only;
+
 	/* We attempt to distribute budget to each Rx queue fairly, but don't
 	 * allow the budget to go below 1 because that would exit polling early.
 	 */
 	budget_per_ring = max(budget/q_vector->num_ringpairs, 1);
 
 	i40e_for_each_ring(ring, q_vector->rx) {
+		int cleaned;
+
 		if (ring_is_ps_enabled(ring))
 			cleaned = i40e_clean_rx_irq_ps(ring, budget_per_ring);
 		else
 			cleaned = i40e_clean_rx_irq_1buf(ring, budget_per_ring);
+
+		work_done += cleaned;
 		/* if we didn't clean as many as budgeted, we must be done */
 		clean_complete &= (budget_per_ring != cleaned);
 	}
 
 	/* If work not completed, return budget and polling will return */
 	if (!clean_complete) {
+tx_only:
 		if (arm_wb)
-			i40e_force_wb(vsi, q_vector);
+			i40evf_force_wb(vsi, q_vector);
 		return budget;
 	}
 
@@ -1393,7 +1373,7 @@ int i40evf_napi_poll(struct napi_struct *napi, int budget)
 		q_vector->arm_wb_state = false;
 
 	/* Work is done so exit the polling mode and re-enable the interrupt */
-	napi_complete(napi);
+	napi_complete_done(napi, work_done);
 	i40e_update_enable_itr(vsi, q_vector);
 	return 0;
 }
@@ -1437,6 +1417,7 @@ static inline int i40evf_tx_prepare_vlan_flags(struct sk_buff *skb,
 	/* else if it is a SW VLAN, check the next protocol and store the tag */
 	} else if (protocol == htons(ETH_P_8021Q)) {
 		struct vlan_hdr *vhdr, _vhdr;
+
 		vhdr = skb_header_pointer(skb, ETH_HLEN, sizeof(_vhdr), &_vhdr);
 		if (!vhdr)
 			return -EINVAL;
@@ -1979,6 +1960,7 @@ static netdev_tx_t i40e_xmit_frame_ring(struct sk_buff *skb,
 	u32 td_cmd = 0;
 	u8 hdr_len = 0;
 	int tso;
+
 	if (0 == i40evf_xmit_descriptor_count(skb, tx_ring))
 		return NETDEV_TX_BUSY;
 
@@ -2006,10 +1988,11 @@ static netdev_tx_t i40e_xmit_frame_ring(struct sk_buff *skb,
 	else if (tso)
 		tx_flags |= I40E_TX_FLAGS_TSO;
 
-	if (i40e_chk_linearize(skb, tx_flags))
+	if (i40e_chk_linearize(skb, tx_flags)) {
 		if (skb_linearize(skb))
 			goto out_drop;
-
+		tx_ring->tx_stats.tx_linearize++;
+	}
 	skb_tx_timestamp(skb);
 
 	/* always enable CRC insertion offload */