[net-next 00/12][pull request] 40GbE Intel Wired LAN Driver Updates 2018-02-12

[net-next 00/12][pull request] 40GbE Intel Wired LAN Driver Updates 2018-02-12

[Jeff Kirsher]

This series contains updates to i40e and i40evf.

Alan fixes a spelling mistake in code comments.  Fixes an issue on older
firmware versions or NPAR enabled PFs which do not support the
I40E_FLAG_DISABLE_FW_LLDP flag and would get into a situation where any
attempt to change any priv flag would be forbidden.

Alex got busy with the ITR code and made several cleanups and fixes so
that we can more easily understand what is going on.  The fixes included
a computational fix when determining the register offset, as well as a
fix for unnecessarily toggling the CLEARPBA bit which could lead to
potential lost events if auto-masking is not enabled.

Filip adds a necessary delay to recover after a EMP reset when using
firmware version 4.33.

Paweł adds a warning message for MFP devices when the link-down-on-close
flag is set because it may affect other partitions.

The following are changes since commit 9a61df9e5f7471fe5be3e02bd0bed726b2761a54:
  Merge tag 'kbuild-v4.16-2' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild
and are available in the git repository at:
  git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue 40GbE

Alan Brady (2):
  i40e: fix typo in function description
  i40e: use changed_flags to check I40E_FLAG_DISABLE_FW_LLDP

Alexander Duyck (8):
  i40e/i40evf: Only track one ITR setting per ring instead of Tx/Rx
  i40e/i40evf: Clean up logic for adaptive ITR
  i40e/i40evf: Clean-up of bits related to using q_vector->reg_idx
  i40e/i40evf: Don't bother setting the CLEARPBA bit
  i40e/i40evf: Use usec value instead of reg value for ITR defines
  i40evf: Correctly populate rxitr_idx and txitr_idx
  i40e/i40evf: Split container ITR into current_itr and target_itr
  i40e/i40evf: Add support for new mechanism of updating adaptive ITR

Filip Sadowski (1):
  i40e: Add delay after EMP reset for firmware to recover

Paweł Jabłoński (1):
  i40e: Warn when setting link-down-on-close while in MFP

 drivers/net/ethernet/intel/i40e/i40e.h             |   3 +-
 drivers/net/ethernet/intel/i40e/i40e_debugfs.c     |  12 +-
 drivers/net/ethernet/intel/i40e/i40e_ethtool.c     |  81 ++--
 drivers/net/ethernet/intel/i40e/i40e_main.c        |  54 ++-
 drivers/net/ethernet/intel/i40e/i40e_txrx.c        | 421 +++++++++++++-------
 drivers/net/ethernet/intel/i40e/i40e_txrx.h        |  68 ++--
 drivers/net/ethernet/intel/i40evf/i40e_txrx.c      | 425 ++++++++++++++-------
 drivers/net/ethernet/intel/i40evf/i40e_txrx.h      |  67 ++--
 drivers/net/ethernet/intel/i40evf/i40evf.h         |   3 +-
 drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c |  46 +--
 drivers/net/ethernet/intel/i40evf/i40evf_main.c    |  22 +-
 .../net/ethernet/intel/i40evf/i40evf_virtchnl.c    |  24 +-
 12 files changed, 769 insertions(+), 457 deletions(-)

-- 
2.14.3

[net-next 01/12] i40e: fix typo in function description

[Jeff Kirsher]

From: Alan Brady <alan.brady@intel.com>

'bufer' should be 'buffer'

Signed-off-by: Alan Brady <alan.brady@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e_txrx.c   | 2 +-
 drivers/net/ethernet/intel/i40evf/i40e_txrx.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
index e554aa6cf070..ddd4b3046046 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
@@ -1991,7 +1991,7 @@ static struct sk_buff *i40e_build_skb(struct i40e_ring *rx_ring,
  * @rx_buffer: rx buffer to pull data from
  *
  * This function will clean up the contents of the rx_buffer.  It will
- * either recycle the bufer or unmap it and free the associated resources.
+ * either recycle the buffer or unmap it and free the associated resources.
  */
 static void i40e_put_rx_buffer(struct i40e_ring *rx_ring,
 			       struct i40e_rx_buffer *rx_buffer)
diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
index 357d6051281f..464200f48d91 100644
--- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
@@ -1273,7 +1273,7 @@ static struct sk_buff *i40e_build_skb(struct i40e_ring *rx_ring,
  * @rx_buffer: rx buffer to pull data from
  *
  * This function will clean up the contents of the rx_buffer.  It will
- * either recycle the bufer or unmap it and free the associated resources.
+ * either recycle the buffer or unmap it and free the associated resources.
  */
 static void i40e_put_rx_buffer(struct i40e_ring *rx_ring,
 			       struct i40e_rx_buffer *rx_buffer)
-- 
2.14.3

[net-next 02/12] i40e/i40evf: Only track one ITR setting per ring instead of Tx/Rx

[Jeff Kirsher]

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

The rings are already split out into Tx and Rx rings so it doesn't make
sense to have any single ring store both a Tx and Rx itr_setting value.
Since that is the case drop the pair in favor of storing just a single ITR
value.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e_debugfs.c     | 12 ++++----
 drivers/net/ethernet/intel/i40e/i40e_ethtool.c     | 32 +++++++++++-----------
 drivers/net/ethernet/intel/i40e/i40e_main.c        | 14 +++++-----
 drivers/net/ethernet/intel/i40e/i40e_txrx.c        |  6 ++--
 drivers/net/ethernet/intel/i40e/i40e_txrx.h        |  3 +-
 drivers/net/ethernet/intel/i40evf/i40e_txrx.c      |  6 ++--
 drivers/net/ethernet/intel/i40evf/i40e_txrx.h      |  3 +-
 drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c | 24 ++++++++--------
 drivers/net/ethernet/intel/i40evf/i40evf_main.c    |  8 +++---
 9 files changed, 53 insertions(+), 55 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_debugfs.c b/drivers/net/ethernet/intel/i40e/i40e_debugfs.c
index 4c3b4243cf65..e9fc51bd6c95 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_debugfs.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_debugfs.c
@@ -315,9 +315,9 @@ static void i40e_dbg_dump_vsi_seid(struct i40e_pf *pf, int seid)
 			 i, rx_ring->vsi,
 			 rx_ring->q_vector);
 		dev_info(&pf->pdev->dev,
-			 "    rx_rings[%i]: rx_itr_setting = %d (%s)\n",
-			 i, rx_ring->rx_itr_setting,
-			 ITR_IS_DYNAMIC(rx_ring->rx_itr_setting) ? "dynamic" : "fixed");
+			 "    rx_rings[%i]: itr_setting = %d (%s)\n",
+			 i, rx_ring->itr_setting,
+			 ITR_IS_DYNAMIC(rx_ring->itr_setting) ? "dynamic" : "fixed");
 	}
 	for (i = 0; i < vsi->num_queue_pairs; i++) {
 		struct i40e_ring *tx_ring = READ_ONCE(vsi->tx_rings[i]);
@@ -366,9 +366,9 @@ static void i40e_dbg_dump_vsi_seid(struct i40e_pf *pf, int seid)
 			 "    tx_rings[%i]: DCB tc = %d\n",
 			 i, tx_ring->dcb_tc);
 		dev_info(&pf->pdev->dev,
-			 "    tx_rings[%i]: tx_itr_setting = %d (%s)\n",
-			 i, tx_ring->tx_itr_setting,
-			 ITR_IS_DYNAMIC(tx_ring->tx_itr_setting) ? "dynamic" : "fixed");
+			 "    tx_rings[%i]: itr_setting = %d (%s)\n",
+			 i, tx_ring->itr_setting,
+			 ITR_IS_DYNAMIC(tx_ring->itr_setting) ? "dynamic" : "fixed");
 	}
 	rcu_read_unlock();
 	dev_info(&pf->pdev->dev,
diff --git a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
index 2f5bee713fef..0fb60336e64b 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
@@ -2244,14 +2244,14 @@ static int __i40e_get_coalesce(struct net_device *netdev,
 	rx_ring = vsi->rx_rings[queue];
 	tx_ring = vsi->tx_rings[queue];
 
-	if (ITR_IS_DYNAMIC(rx_ring->rx_itr_setting))
+	if (ITR_IS_DYNAMIC(rx_ring->itr_setting))
 		ec->use_adaptive_rx_coalesce = 1;
 
-	if (ITR_IS_DYNAMIC(tx_ring->tx_itr_setting))
+	if (ITR_IS_DYNAMIC(tx_ring->itr_setting))
 		ec->use_adaptive_tx_coalesce = 1;
 
-	ec->rx_coalesce_usecs = rx_ring->rx_itr_setting & ~I40E_ITR_DYNAMIC;
-	ec->tx_coalesce_usecs = tx_ring->tx_itr_setting & ~I40E_ITR_DYNAMIC;
+	ec->rx_coalesce_usecs = rx_ring->itr_setting & ~I40E_ITR_DYNAMIC;
+	ec->tx_coalesce_usecs = tx_ring->itr_setting & ~I40E_ITR_DYNAMIC;
 
 	/* we use the _usecs_high to store/set the interrupt rate limit
 	 * that the hardware supports, that almost but not quite
@@ -2315,26 +2315,26 @@ static void i40e_set_itr_per_queue(struct i40e_vsi *vsi,
 
 	intrl = i40e_intrl_usec_to_reg(vsi->int_rate_limit);
 
-	rx_ring->rx_itr_setting = ec->rx_coalesce_usecs;
-	tx_ring->tx_itr_setting = ec->tx_coalesce_usecs;
+	rx_ring->itr_setting = ec->rx_coalesce_usecs;
+	tx_ring->itr_setting = ec->tx_coalesce_usecs;
 
 	if (ec->use_adaptive_rx_coalesce)
-		rx_ring->rx_itr_setting |= I40E_ITR_DYNAMIC;
+		rx_ring->itr_setting |= I40E_ITR_DYNAMIC;
 	else
-		rx_ring->rx_itr_setting &= ~I40E_ITR_DYNAMIC;
+		rx_ring->itr_setting &= ~I40E_ITR_DYNAMIC;
 
 	if (ec->use_adaptive_tx_coalesce)
-		tx_ring->tx_itr_setting |= I40E_ITR_DYNAMIC;
+		tx_ring->itr_setting |= I40E_ITR_DYNAMIC;
 	else
-		tx_ring->tx_itr_setting &= ~I40E_ITR_DYNAMIC;
+		tx_ring->itr_setting &= ~I40E_ITR_DYNAMIC;
 
 	q_vector = rx_ring->q_vector;
-	q_vector->rx.itr = ITR_TO_REG(rx_ring->rx_itr_setting);
+	q_vector->rx.itr = ITR_TO_REG(rx_ring->itr_setting);
 	vector = vsi->base_vector + q_vector->v_idx;
 	wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1), q_vector->rx.itr);
 
 	q_vector = tx_ring->q_vector;
-	q_vector->tx.itr = ITR_TO_REG(tx_ring->tx_itr_setting);
+	q_vector->tx.itr = ITR_TO_REG(tx_ring->itr_setting);
 	vector = vsi->base_vector + q_vector->v_idx;
 	wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1), q_vector->tx.itr);
 
@@ -2364,11 +2364,11 @@ static int __i40e_set_coalesce(struct net_device *netdev,
 		vsi->work_limit = ec->tx_max_coalesced_frames_irq;
 
 	if (queue < 0) {
-		cur_rx_itr = vsi->rx_rings[0]->rx_itr_setting;
-		cur_tx_itr = vsi->tx_rings[0]->tx_itr_setting;
+		cur_rx_itr = vsi->rx_rings[0]->itr_setting;
+		cur_tx_itr = vsi->tx_rings[0]->itr_setting;
 	} else if (queue < vsi->num_queue_pairs) {
-		cur_rx_itr = vsi->rx_rings[queue]->rx_itr_setting;
-		cur_tx_itr = vsi->tx_rings[queue]->tx_itr_setting;
+		cur_rx_itr = vsi->rx_rings[queue]->itr_setting;
+		cur_tx_itr = vsi->tx_rings[queue]->itr_setting;
 	} else {
 		netif_info(pf, drv, netdev, "Invalid queue value, queue range is 0 - %d\n",
 			   vsi->num_queue_pairs - 1);
diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c
index e31adbc75f9c..005a6d6a20fb 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_main.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_main.c
@@ -3450,11 +3450,11 @@ static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
 		struct i40e_q_vector *q_vector = vsi->q_vectors[i];
 
 		q_vector->itr_countdown = ITR_COUNTDOWN_START;
-		q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[i]->rx_itr_setting);
+		q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[i]->itr_setting);
 		q_vector->rx.latency_range = I40E_LOW_LATENCY;
 		wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
 		     q_vector->rx.itr);
-		q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[i]->tx_itr_setting);
+		q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[i]->itr_setting);
 		q_vector->tx.latency_range = I40E_LOW_LATENCY;
 		wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
 		     q_vector->tx.itr);
@@ -3559,10 +3559,10 @@ static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
 
 	/* set the ITR configuration */
 	q_vector->itr_countdown = ITR_COUNTDOWN_START;
-	q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[0]->rx_itr_setting);
+	q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[0]->itr_setting);
 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
-	q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[0]->tx_itr_setting);
+	q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[0]->itr_setting);
 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
 	wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
 
@@ -10018,7 +10018,7 @@ static int i40e_alloc_rings(struct i40e_vsi *vsi)
 		ring->dcb_tc = 0;
 		if (vsi->back->hw_features & I40E_HW_WB_ON_ITR_CAPABLE)
 			ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
-		ring->tx_itr_setting = pf->tx_itr_default;
+		ring->itr_setting = pf->tx_itr_default;
 		vsi->tx_rings[i] = ring++;
 
 		if (!i40e_enabled_xdp_vsi(vsi))
@@ -10036,7 +10036,7 @@ static int i40e_alloc_rings(struct i40e_vsi *vsi)
 		if (vsi->back->hw_features & I40E_HW_WB_ON_ITR_CAPABLE)
 			ring->flags = I40E_TXR_FLAGS_WB_ON_ITR;
 		set_ring_xdp(ring);
-		ring->tx_itr_setting = pf->tx_itr_default;
+		ring->itr_setting = pf->tx_itr_default;
 		vsi->xdp_rings[i] = ring++;
 
 setup_rx:
@@ -10049,7 +10049,7 @@ static int i40e_alloc_rings(struct i40e_vsi *vsi)
 		ring->count = vsi->num_desc;
 		ring->size = 0;
 		ring->dcb_tc = 0;
-		ring->rx_itr_setting = pf->rx_itr_default;
+		ring->itr_setting = pf->rx_itr_default;
 		vsi->rx_rings[i] = ring;
 	}
 
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
index ddd4b3046046..b9121a87ee46 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
@@ -2290,12 +2290,12 @@ static u32 i40e_buildreg_itr(const int type, const u16 itr)
 #define INTREG I40E_PFINT_DYN_CTLN
 static inline int get_rx_itr(struct i40e_vsi *vsi, int idx)
 {
-	return vsi->rx_rings[idx]->rx_itr_setting;
+	return vsi->rx_rings[idx]->itr_setting;
 }
 
 static inline int get_tx_itr(struct i40e_vsi *vsi, int idx)
 {
-	return vsi->tx_rings[idx]->tx_itr_setting;
+	return vsi->tx_rings[idx]->itr_setting;
 }
 
 /**
@@ -2322,7 +2322,7 @@ static inline void i40e_update_enable_itr(struct i40e_vsi *vsi,
 	/* avoid dynamic calculation if in countdown mode OR if
 	 * all dynamic is disabled
 	 */
-	rxval = txval = i40e_buildreg_itr(I40E_ITR_NONE, 0);
+	txval = i40e_buildreg_itr(I40E_ITR_NONE, 0);
 
 	rx_itr_setting = get_rx_itr(vsi, idx);
 	tx_itr_setting = get_tx_itr(vsi, idx);
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.h b/drivers/net/ethernet/intel/i40e/i40e_txrx.h
index 701b708628b0..1758dd3bf91b 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.h
@@ -382,8 +382,7 @@ struct i40e_ring {
 	 * these values always store the USER setting, and must be converted
 	 * before programming to a register.
 	 */
-	u16 rx_itr_setting;
-	u16 tx_itr_setting;
+	u16 itr_setting;
 
 	u16 count;			/* Number of descriptors */
 	u16 reg_idx;			/* HW register index of the ring */
diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
index 464200f48d91..291130af2985 100644
--- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
@@ -1475,14 +1475,14 @@ static inline int get_rx_itr(struct i40e_vsi *vsi, int idx)
 {
 	struct i40evf_adapter *adapter = vsi->back;
 
-	return adapter->rx_rings[idx].rx_itr_setting;
+	return adapter->rx_rings[idx].itr_setting;
 }
 
 static inline int get_tx_itr(struct i40e_vsi *vsi, int idx)
 {
 	struct i40evf_adapter *adapter = vsi->back;
 
-	return adapter->tx_rings[idx].tx_itr_setting;
+	return adapter->tx_rings[idx].itr_setting;
 }
 
 /**
@@ -1503,7 +1503,7 @@ static inline void i40e_update_enable_itr(struct i40e_vsi *vsi,
 	/* avoid dynamic calculation if in countdown mode OR if
 	 * all dynamic is disabled
 	 */
-	rxval = txval = i40e_buildreg_itr(I40E_ITR_NONE, 0);
+	txval = i40e_buildreg_itr(I40E_ITR_NONE, 0);
 
 	rx_itr_setting = get_rx_itr(vsi, idx);
 	tx_itr_setting = get_tx_itr(vsi, idx);
diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.h b/drivers/net/ethernet/intel/i40evf/i40e_txrx.h
index 7798a6645c3f..038ed0e2acb7 100644
--- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.h
+++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.h
@@ -362,8 +362,7 @@ struct i40e_ring {
 	 * these values always store the USER setting, and must be converted
 	 * before programming to a register.
 	 */
-	u16 rx_itr_setting;
-	u16 tx_itr_setting;
+	u16 itr_setting;
 
 	u16 count;			/* Number of descriptors */
 	u16 reg_idx;			/* HW register index of the ring */
diff --git a/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c b/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c
index e2d8aa19d205..11dfdc882934 100644
--- a/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c
+++ b/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c
@@ -457,14 +457,14 @@ static int __i40evf_get_coalesce(struct net_device *netdev,
 	rx_ring = &adapter->rx_rings[queue];
 	tx_ring = &adapter->tx_rings[queue];
 
-	if (ITR_IS_DYNAMIC(rx_ring->rx_itr_setting))
+	if (ITR_IS_DYNAMIC(rx_ring->itr_setting))
 		ec->use_adaptive_rx_coalesce = 1;
 
-	if (ITR_IS_DYNAMIC(tx_ring->tx_itr_setting))
+	if (ITR_IS_DYNAMIC(tx_ring->itr_setting))
 		ec->use_adaptive_tx_coalesce = 1;
 
-	ec->rx_coalesce_usecs = rx_ring->rx_itr_setting & ~I40E_ITR_DYNAMIC;
-	ec->tx_coalesce_usecs = tx_ring->tx_itr_setting & ~I40E_ITR_DYNAMIC;
+	ec->rx_coalesce_usecs = rx_ring->itr_setting & ~I40E_ITR_DYNAMIC;
+	ec->tx_coalesce_usecs = tx_ring->itr_setting & ~I40E_ITR_DYNAMIC;
 
 	return 0;
 }
@@ -519,24 +519,24 @@ static void i40evf_set_itr_per_queue(struct i40evf_adapter *adapter,
 	struct i40e_q_vector *q_vector;
 	u16 vector;
 
-	rx_ring->rx_itr_setting = ec->rx_coalesce_usecs;
-	tx_ring->tx_itr_setting = ec->tx_coalesce_usecs;
+	rx_ring->itr_setting = ec->rx_coalesce_usecs;
+	tx_ring->itr_setting = ec->tx_coalesce_usecs;
 
-	rx_ring->rx_itr_setting |= I40E_ITR_DYNAMIC;
+	rx_ring->itr_setting |= I40E_ITR_DYNAMIC;
 	if (!ec->use_adaptive_rx_coalesce)
-		rx_ring->rx_itr_setting ^= I40E_ITR_DYNAMIC;
+		rx_ring->itr_setting ^= I40E_ITR_DYNAMIC;
 
-	tx_ring->tx_itr_setting |= I40E_ITR_DYNAMIC;
+	tx_ring->itr_setting |= I40E_ITR_DYNAMIC;
 	if (!ec->use_adaptive_tx_coalesce)
-		tx_ring->tx_itr_setting ^= I40E_ITR_DYNAMIC;
+		tx_ring->itr_setting ^= I40E_ITR_DYNAMIC;
 
 	q_vector = rx_ring->q_vector;
-	q_vector->rx.itr = ITR_TO_REG(rx_ring->rx_itr_setting);
+	q_vector->rx.itr = ITR_TO_REG(rx_ring->itr_setting);
 	vector = vsi->base_vector + q_vector->v_idx;
 	wr32(hw, I40E_VFINT_ITRN1(I40E_RX_ITR, vector - 1), q_vector->rx.itr);
 
 	q_vector = tx_ring->q_vector;
-	q_vector->tx.itr = ITR_TO_REG(tx_ring->tx_itr_setting);
+	q_vector->tx.itr = ITR_TO_REG(tx_ring->itr_setting);
 	vector = vsi->base_vector + q_vector->v_idx;
 	wr32(hw, I40E_VFINT_ITRN1(I40E_TX_ITR, vector - 1), q_vector->tx.itr);
 
diff --git a/drivers/net/ethernet/intel/i40evf/i40evf_main.c b/drivers/net/ethernet/intel/i40evf/i40evf_main.c
index 16989ad2ca90..a5fb540c2637 100644
--- a/drivers/net/ethernet/intel/i40evf/i40evf_main.c
+++ b/drivers/net/ethernet/intel/i40evf/i40evf_main.c
@@ -354,7 +354,7 @@ i40evf_map_vector_to_rxq(struct i40evf_adapter *adapter, int v_idx, int r_idx)
 	q_vector->rx.ring = rx_ring;
 	q_vector->rx.count++;
 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
-	q_vector->rx.itr = ITR_TO_REG(rx_ring->rx_itr_setting);
+	q_vector->rx.itr = ITR_TO_REG(rx_ring->itr_setting);
 	q_vector->ring_mask |= BIT(r_idx);
 	q_vector->itr_countdown = ITR_COUNTDOWN_START;
 	wr32(hw, I40E_VFINT_ITRN1(I40E_RX_ITR, v_idx - 1), q_vector->rx.itr);
@@ -379,7 +379,7 @@ i40evf_map_vector_to_txq(struct i40evf_adapter *adapter, int v_idx, int t_idx)
 	q_vector->tx.ring = tx_ring;
 	q_vector->tx.count++;
 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
-	q_vector->tx.itr = ITR_TO_REG(tx_ring->tx_itr_setting);
+	q_vector->tx.itr = ITR_TO_REG(tx_ring->itr_setting);
 	q_vector->itr_countdown = ITR_COUNTDOWN_START;
 	q_vector->num_ringpairs++;
 	wr32(hw, I40E_VFINT_ITRN1(I40E_TX_ITR, v_idx - 1), q_vector->tx.itr);
@@ -1169,7 +1169,7 @@ static int i40evf_alloc_queues(struct i40evf_adapter *adapter)
 		tx_ring->netdev = adapter->netdev;
 		tx_ring->dev = &adapter->pdev->dev;
 		tx_ring->count = adapter->tx_desc_count;
-		tx_ring->tx_itr_setting = I40E_ITR_TX_DEF;
+		tx_ring->itr_setting = I40E_ITR_TX_DEF;
 		if (adapter->flags & I40EVF_FLAG_WB_ON_ITR_CAPABLE)
 			tx_ring->flags |= I40E_TXR_FLAGS_WB_ON_ITR;
 
@@ -1178,7 +1178,7 @@ static int i40evf_alloc_queues(struct i40evf_adapter *adapter)
 		rx_ring->netdev = adapter->netdev;
 		rx_ring->dev = &adapter->pdev->dev;
 		rx_ring->count = adapter->rx_desc_count;
-		rx_ring->rx_itr_setting = I40E_ITR_RX_DEF;
+		rx_ring->itr_setting = I40E_ITR_RX_DEF;
 	}
 
 	adapter->num_active_queues = num_active_queues;
-- 
2.14.3

[net-next 03/12] i40e/i40evf: Clean up logic for adaptive ITR

[Jeff Kirsher]

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

The logic for dynamic ITR update is confusing at best as there were odd
paths chosen for how to find the rings associated with a given queue based
on the vector index and other inconsistencies throughout the code.

This patch is an attempt to clean up the logic so that we can more easily
understand what is going on. Specifically if there is a Rx or Tx ring that
is enabled in dynamic mode on the q_vector it is allowed to override the
other side of the interrupt moderation. While it isn't correct all this
patch is doing is cleaning up the logic for now so that when we come
through and fix it we can more easily identify that this is wrong.

The other big change made here is that we replace references to:
	vsi->rx_rings[q_vector->v_idx]->itr_setting
with:
	q_vector->rx.ring->itr_setting

The general idea is we can avoid the long pointer chase since just
accessing q_vector->rx.ring is a single pointer access versus having to
chase down vsi->rx_rings, and then finding the pointer in the array, and
finally chasing down the itr_setting from there.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e_txrx.c   | 55 +++++++------------------
 drivers/net/ethernet/intel/i40evf/i40e_txrx.c | 59 +++++++--------------------
 2 files changed, 28 insertions(+), 86 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
index b9121a87ee46..6b3fe5d26dc1 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
@@ -1016,6 +1016,9 @@ static bool i40e_set_new_dynamic_itr(struct i40e_ring_container *rc)
 	int bytes_per_usec;
 	unsigned int usecs, estimated_usecs;
 
+	if (!rc->ring || !ITR_IS_DYNAMIC(rc->ring->itr_setting))
+		return false;
+
 	if (rc->total_packets == 0 || !rc->itr)
 		return false;
 
@@ -2288,15 +2291,6 @@ static u32 i40e_buildreg_itr(const int type, const u16 itr)
 
 /* a small macro to shorten up some long lines */
 #define INTREG I40E_PFINT_DYN_CTLN
-static inline int get_rx_itr(struct i40e_vsi *vsi, int idx)
-{
-	return vsi->rx_rings[idx]->itr_setting;
-}
-
-static inline int get_tx_itr(struct i40e_vsi *vsi, int idx)
-{
-	return vsi->tx_rings[idx]->itr_setting;
-}
 
 /**
  * i40e_update_enable_itr - Update itr and re-enable MSIX interrupt
@@ -2309,9 +2303,7 @@ static inline void i40e_update_enable_itr(struct i40e_vsi *vsi,
 {
 	struct i40e_hw *hw = &vsi->back->hw;
 	bool rx = false, tx = false;
-	u32 rxval, txval;
-	int idx = q_vector->v_idx;
-	int rx_itr_setting, tx_itr_setting;
+	u32 txval;
 
 	/* If we don't have MSIX, then we only need to re-enable icr0 */
 	if (!(vsi->back->flags & I40E_FLAG_MSIX_ENABLED)) {
@@ -2319,29 +2311,15 @@ static inline void i40e_update_enable_itr(struct i40e_vsi *vsi,
 		return;
 	}
 
-	/* avoid dynamic calculation if in countdown mode OR if
-	 * all dynamic is disabled
-	 */
 	txval = i40e_buildreg_itr(I40E_ITR_NONE, 0);
 
-	rx_itr_setting = get_rx_itr(vsi, idx);
-	tx_itr_setting = get_tx_itr(vsi, idx);
-
-	if (q_vector->itr_countdown > 0 ||
-	    (!ITR_IS_DYNAMIC(rx_itr_setting) &&
-	     !ITR_IS_DYNAMIC(tx_itr_setting))) {
+	/* avoid dynamic calculation if in countdown mode */
+	if (q_vector->itr_countdown > 0)
 		goto enable_int;
-	}
 
-	if (ITR_IS_DYNAMIC(rx_itr_setting)) {
-		rx = i40e_set_new_dynamic_itr(&q_vector->rx);
-		rxval = i40e_buildreg_itr(I40E_RX_ITR, q_vector->rx.itr);
-	}
-
-	if (ITR_IS_DYNAMIC(tx_itr_setting)) {
-		tx = i40e_set_new_dynamic_itr(&q_vector->tx);
-		txval = i40e_buildreg_itr(I40E_TX_ITR, q_vector->tx.itr);
-	}
+	/* these will return false if dynamic mode is disabled */
+	rx = i40e_set_new_dynamic_itr(&q_vector->rx);
+	tx = i40e_set_new_dynamic_itr(&q_vector->tx);
 
 	if (rx || tx) {
 		/* get the higher of the two ITR adjustments and
@@ -2349,25 +2327,20 @@ static inline void i40e_update_enable_itr(struct i40e_vsi *vsi,
 		 * when in adaptive mode (Rx and/or Tx)
 		 */
 		u16 itr = max(q_vector->tx.itr, q_vector->rx.itr);
+		u32 rxval;
 
 		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);
+		rxval = i40e_buildreg_itr(I40E_RX_ITR, itr) | BIT(31);
+
 		/* don't check _DOWN because interrupt isn't being enabled */
 		wr32(hw, INTREG(q_vector->reg_idx), rxval);
+
+		txval = i40e_buildreg_itr(I40E_TX_ITR, itr);
 	}
 
 enable_int:
diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
index 291130af2985..3fd7e9731f49 100644
--- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
@@ -413,6 +413,9 @@ static bool i40e_set_new_dynamic_itr(struct i40e_ring_container *rc)
 	int bytes_per_usec;
 	unsigned int usecs, estimated_usecs;
 
+	if (!rc->ring || !ITR_IS_DYNAMIC(rc->ring->itr_setting))
+		return false;
+
 	if (rc->total_packets == 0 || !rc->itr)
 		return false;
 
@@ -1471,19 +1474,6 @@ static u32 i40e_buildreg_itr(const int type, const u16 itr)
 
 /* a small macro to shorten up some long lines */
 #define INTREG I40E_VFINT_DYN_CTLN1
-static inline int get_rx_itr(struct i40e_vsi *vsi, int idx)
-{
-	struct i40evf_adapter *adapter = vsi->back;
-
-	return adapter->rx_rings[idx].itr_setting;
-}
-
-static inline int get_tx_itr(struct i40e_vsi *vsi, int idx)
-{
-	struct i40evf_adapter *adapter = vsi->back;
-
-	return adapter->tx_rings[idx].itr_setting;
-}
 
 /**
  * i40e_update_enable_itr - Update itr and re-enable MSIX interrupt
@@ -1496,33 +1486,17 @@ static inline void i40e_update_enable_itr(struct i40e_vsi *vsi,
 {
 	struct i40e_hw *hw = &vsi->back->hw;
 	bool rx = false, tx = false;
-	u32 rxval, txval;
-	int idx = q_vector->v_idx;
-	int rx_itr_setting, tx_itr_setting;
+	u32 txval;
 
-	/* avoid dynamic calculation if in countdown mode OR if
-	 * all dynamic is disabled
-	 */
 	txval = i40e_buildreg_itr(I40E_ITR_NONE, 0);
 
-	rx_itr_setting = get_rx_itr(vsi, idx);
-	tx_itr_setting = get_tx_itr(vsi, idx);
-
-	if (q_vector->itr_countdown > 0 ||
-	    (!ITR_IS_DYNAMIC(rx_itr_setting) &&
-	     !ITR_IS_DYNAMIC(tx_itr_setting))) {
+	/* avoid dynamic calculation if in countdown mode */
+	if (q_vector->itr_countdown > 0)
 		goto enable_int;
-	}
 
-	if (ITR_IS_DYNAMIC(rx_itr_setting)) {
-		rx = i40e_set_new_dynamic_itr(&q_vector->rx);
-		rxval = i40e_buildreg_itr(I40E_RX_ITR, q_vector->rx.itr);
-	}
-
-	if (ITR_IS_DYNAMIC(tx_itr_setting)) {
-		tx = i40e_set_new_dynamic_itr(&q_vector->tx);
-		txval = i40e_buildreg_itr(I40E_TX_ITR, q_vector->tx.itr);
-	}
+	/* these will return false if dynamic mode is disabled */
+	rx = i40e_set_new_dynamic_itr(&q_vector->rx);
+	tx = i40e_set_new_dynamic_itr(&q_vector->tx);
 
 	if (rx || tx) {
 		/* get the higher of the two ITR adjustments and
@@ -1530,25 +1504,20 @@ static inline void i40e_update_enable_itr(struct i40e_vsi *vsi,
 		 * when in adaptive mode (Rx and/or Tx)
 		 */
 		u16 itr = max(q_vector->tx.itr, q_vector->rx.itr);
+		u32 rxval;
 
 		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);
+		rxval = i40e_buildreg_itr(I40E_RX_ITR, itr) | BIT(31);
+
 		/* don't check _DOWN because interrupt isn't being enabled */
 		wr32(hw, INTREG(q_vector->reg_idx), rxval);
+
+		txval = i40e_buildreg_itr(I40E_TX_ITR, itr);
 	}
 
 enable_int:
-- 
2.14.3

[net-next 04/12] i40e: Add delay after EMP reset for firmware to recover

[Jeff Kirsher]

From: Filip Sadowski <filip.sadowski@intel.com>

This patch adds necessary delay for 4.33 firmware to recover after
EMP reset. Without this patch driver occasionally reinitializes
structures too quickly to communicate with firmware after EMP reset
causing AdminQ to timeout.

Signed-off-by: Filip Sadowski <filip.sadowski@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e_main.c | 11 +++++++++++
 1 file changed, 11 insertions(+)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c
index 005a6d6a20fb..a88fdb8bf5f0 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_main.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_main.c
@@ -9215,6 +9215,17 @@ static void i40e_rebuild(struct i40e_pf *pf, bool reinit, bool lock_acquired)
 	}
 	i40e_get_oem_version(&pf->hw);
 
+	if (test_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state) &&
+	    ((hw->aq.fw_maj_ver == 4 && hw->aq.fw_min_ver <= 33) ||
+	     hw->aq.fw_maj_ver < 4) && hw->mac.type == I40E_MAC_XL710) {
+		/* The following delay is necessary for 4.33 firmware and older
+		 * to recover after EMP reset. 200 ms should suffice but we
+		 * put here 300 ms to be sure that FW is ready to operate
+		 * after reset.
+		 */
+		mdelay(300);
+	}
+
 	/* re-verify the eeprom if we just had an EMP reset */
 	if (test_and_clear_bit(__I40E_EMP_RESET_INTR_RECEIVED, pf->state))
 		i40e_verify_eeprom(pf);
-- 
2.14.3

[net-next 05/12] i40e: Warn when setting link-down-on-close while in MFP

[Jeff Kirsher]

From: Paweł Jabłoński <pawel.jablonski@intel.com>

This patch adds a warning message when the link-down-on-close flag is
setting on. The warning is printed only on MFP devices

Signed-off-by: Paweł Jabłoński <pawel.jablonski@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e_ethtool.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
index 0fb60336e64b..5ca63c5d36b4 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
@@ -4479,6 +4479,12 @@ static int i40e_set_priv_flags(struct net_device *dev, u32 flags)
 		}
 	}
 
+	if ((changed_flags & pf->flags &
+	     I40E_FLAG_LINK_DOWN_ON_CLOSE_ENABLED) &&
+	    (pf->flags & I40E_FLAG_MFP_ENABLED))
+		dev_warn(&pf->pdev->dev,
+			 "Turning on link-down-on-close flag may affect other partitions\n");
+
 	if (changed_flags & I40E_FLAG_DISABLE_FW_LLDP) {
 		if (pf->flags & I40E_FLAG_DISABLE_FW_LLDP) {
 			struct i40e_dcbx_config *dcbcfg;
-- 
2.14.3

[net-next 06/12] i40e: use changed_flags to check I40E_FLAG_DISABLE_FW_LLDP

[Jeff Kirsher]

From: Alan Brady <alan.brady@intel.com>

Currently in i40e_set_priv_flags we use new_flags to check for the
I40E_FLAG_DISABLE_FW_LLDP flag.  This is an issue for a few a reasons.
DISABLE_FW_LLDP is persistent across reboots/driver reloads.  This means
we need some way to detect if FW LLDP is enabled on init.  We do this by
trying to init_dcb and if it fails with EPERM we know LLDP is disabled
in FW.

This could be a problem on older FW versions or NPAR enabled PFs because
there are situations where the FW could disable LLDP, but they do _not_
support using this flag to change it.  If we do end up in this
situation, the flag will be set, then when the user tries to change any
priv flags, the driver thinks the user is trying to disable FW LLDP on a
FW that doesn't support it and essentially forbids any priv flag
changes.

The fix is simple, instead of checking if this flag is set, we should be
checking if the user is trying to _change_ the flag on unsupported FW
versions.

This patch also adds a comment explaining that the cmpxchg is the point
of no return.  Once we put the new flags into pf->flags we can't back
out.

Signed-off-by: Alan Brady <alan.brady@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e_ethtool.c | 22 +++++++++++++++-------
 1 file changed, 15 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
index 5ca63c5d36b4..5ee27358922a 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
@@ -4406,6 +4406,8 @@ static int i40e_set_priv_flags(struct net_device *dev, u32 flags)
 	}
 
 flags_complete:
+	changed_flags = orig_flags ^ new_flags;
+
 	/* Before we finalize any flag changes, we need to perform some
 	 * checks to ensure that the changes are supported and safe.
 	 */
@@ -4415,13 +4417,17 @@ static int i40e_set_priv_flags(struct net_device *dev, u32 flags)
 	    !(pf->hw_features & I40E_HW_ATR_EVICT_CAPABLE))
 		return -EOPNOTSUPP;
 
-	/* Disable FW LLDP not supported if NPAR active or if FW
-	 * API version < 1.7
+	/* If the driver detected FW LLDP was disabled on init, this flag could
+	 * be set, however we do not support _changing_ the flag if NPAR is
+	 * enabled or FW API version < 1.7.  There are situations where older
+	 * FW versions/NPAR enabled PFs could disable LLDP, however we _must_
+	 * not allow the user to enable/disable LLDP with this flag on
+	 * unsupported FW versions.
 	 */
-	if (new_flags & I40E_FLAG_DISABLE_FW_LLDP) {
+	if (changed_flags & I40E_FLAG_DISABLE_FW_LLDP) {
 		if (pf->hw.func_caps.npar_enable) {
 			dev_warn(&pf->pdev->dev,
-				 "Unable to stop FW LLDP if NPAR active\n");
+				 "Unable to change FW LLDP if NPAR active\n");
 			return -EOPNOTSUPP;
 		}
 
@@ -4429,7 +4435,7 @@ static int i40e_set_priv_flags(struct net_device *dev, u32 flags)
 		    (pf->hw.aq.api_maj_ver == 1 &&
 		     pf->hw.aq.api_min_ver < 7)) {
 			dev_warn(&pf->pdev->dev,
-				 "FW ver does not support stopping FW LLDP\n");
+				 "FW ver does not support changing FW LLDP\n");
 			return -EOPNOTSUPP;
 		}
 	}
@@ -4439,6 +4445,10 @@ static int i40e_set_priv_flags(struct net_device *dev, u32 flags)
 	 * something else has modified the flags variable since we copied it
 	 * originally. We'll just punt with an error and log something in the
 	 * message buffer.
+	 *
+	 * This is the point of no return for this function.  We need to have
+	 * checked any discrepancies or misconfigurations and returned
+	 * EOPNOTSUPP before updating pf->flags here.
 	 */
 	if (cmpxchg64(&pf->flags, orig_flags, new_flags) != orig_flags) {
 		dev_warn(&pf->pdev->dev,
@@ -4446,8 +4456,6 @@ static int i40e_set_priv_flags(struct net_device *dev, u32 flags)
 		return -EAGAIN;
 	}
 
-	changed_flags = orig_flags ^ new_flags;
-
 	/* Process any additional changes needed as a result of flag changes.
 	 * The changed_flags value reflects the list of bits that were
 	 * changed in the code above.
-- 
2.14.3

[net-next 07/12] i40e/i40evf: Clean-up of bits related to using q_vector->reg_idx

[Jeff Kirsher]

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

This patch is a further clean-up related to the change over to using
q_vector->reg_idx when accessing the ITR registers. Specifically the code
appears to have several other spots where we were computing the register
offset manually and this resulted in errors in a few spots.

Specifically in the i40evf functions for mapping queues to vectors it
appears we may have had an off by 1 error since (v_idx - 1) for the first
q_vector with an index of 0 would result in us returning -1 if I am not
mistaken.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e_ethtool.c     | 12 ++++++------
 drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c | 12 +++++-------
 drivers/net/ethernet/intel/i40evf/i40evf_main.c    |  6 ++++--
 3 files changed, 15 insertions(+), 15 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
index 5ee27358922a..8cc9198ac32f 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
@@ -2311,7 +2311,7 @@ static void i40e_set_itr_per_queue(struct i40e_vsi *vsi,
 	struct i40e_pf *pf = vsi->back;
 	struct i40e_hw *hw = &pf->hw;
 	struct i40e_q_vector *q_vector;
-	u16 vector, intrl;
+	u16 intrl;
 
 	intrl = i40e_intrl_usec_to_reg(vsi->int_rate_limit);
 
@@ -2330,15 +2330,15 @@ static void i40e_set_itr_per_queue(struct i40e_vsi *vsi,
 
 	q_vector = rx_ring->q_vector;
 	q_vector->rx.itr = ITR_TO_REG(rx_ring->itr_setting);
-	vector = vsi->base_vector + q_vector->v_idx;
-	wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1), q_vector->rx.itr);
+	wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, q_vector->reg_idx),
+	     q_vector->rx.itr);
 
 	q_vector = tx_ring->q_vector;
 	q_vector->tx.itr = ITR_TO_REG(tx_ring->itr_setting);
-	vector = vsi->base_vector + q_vector->v_idx;
-	wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1), q_vector->tx.itr);
+	wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, q_vector->reg_idx),
+	     q_vector->tx.itr);
 
-	wr32(hw, I40E_PFINT_RATEN(vector - 1), intrl);
+	wr32(hw, I40E_PFINT_RATEN(q_vector->reg_idx), intrl);
 	i40e_flush(hw);
 }
 
diff --git a/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c b/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c
index 11dfdc882934..ed5b8ec4d2a2 100644
--- a/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c
+++ b/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c
@@ -502,7 +502,7 @@ static int i40evf_get_per_queue_coalesce(struct net_device *netdev,
 
 /**
  * i40evf_set_itr_per_queue - set ITR values for specific queue
- * @vsi: the VSI to set values for
+ * @adapter: the VF adapter struct to set values for
  * @ec: coalesce settings from ethtool
  * @queue: the queue to modify
  *
@@ -514,10 +514,8 @@ static void i40evf_set_itr_per_queue(struct i40evf_adapter *adapter,
 {
 	struct i40e_ring *rx_ring = &adapter->rx_rings[queue];
 	struct i40e_ring *tx_ring = &adapter->tx_rings[queue];
-	struct i40e_vsi *vsi = &adapter->vsi;
 	struct i40e_hw *hw = &adapter->hw;
 	struct i40e_q_vector *q_vector;
-	u16 vector;
 
 	rx_ring->itr_setting = ec->rx_coalesce_usecs;
 	tx_ring->itr_setting = ec->tx_coalesce_usecs;
@@ -532,13 +530,13 @@ static void i40evf_set_itr_per_queue(struct i40evf_adapter *adapter,
 
 	q_vector = rx_ring->q_vector;
 	q_vector->rx.itr = ITR_TO_REG(rx_ring->itr_setting);
-	vector = vsi->base_vector + q_vector->v_idx;
-	wr32(hw, I40E_VFINT_ITRN1(I40E_RX_ITR, vector - 1), q_vector->rx.itr);
+	wr32(hw, I40E_VFINT_ITRN1(I40E_RX_ITR, q_vector->reg_idx),
+	     q_vector->rx.itr);
 
 	q_vector = tx_ring->q_vector;
 	q_vector->tx.itr = ITR_TO_REG(tx_ring->itr_setting);
-	vector = vsi->base_vector + q_vector->v_idx;
-	wr32(hw, I40E_VFINT_ITRN1(I40E_TX_ITR, vector - 1), q_vector->tx.itr);
+	wr32(hw, I40E_VFINT_ITRN1(I40E_TX_ITR, q_vector->reg_idx),
+	     q_vector->tx.itr);
 
 	i40e_flush(hw);
 }
diff --git a/drivers/net/ethernet/intel/i40evf/i40evf_main.c b/drivers/net/ethernet/intel/i40evf/i40evf_main.c
index a5fb540c2637..f648e5e97529 100644
--- a/drivers/net/ethernet/intel/i40evf/i40evf_main.c
+++ b/drivers/net/ethernet/intel/i40evf/i40evf_main.c
@@ -357,7 +357,8 @@ i40evf_map_vector_to_rxq(struct i40evf_adapter *adapter, int v_idx, int r_idx)
 	q_vector->rx.itr = ITR_TO_REG(rx_ring->itr_setting);
 	q_vector->ring_mask |= BIT(r_idx);
 	q_vector->itr_countdown = ITR_COUNTDOWN_START;
-	wr32(hw, I40E_VFINT_ITRN1(I40E_RX_ITR, v_idx - 1), q_vector->rx.itr);
+	wr32(hw, I40E_VFINT_ITRN1(I40E_RX_ITR, q_vector->reg_idx),
+	     q_vector->rx.itr);
 }
 
 /**
@@ -382,7 +383,8 @@ i40evf_map_vector_to_txq(struct i40evf_adapter *adapter, int v_idx, int t_idx)
 	q_vector->tx.itr = ITR_TO_REG(tx_ring->itr_setting);
 	q_vector->itr_countdown = ITR_COUNTDOWN_START;
 	q_vector->num_ringpairs++;
-	wr32(hw, I40E_VFINT_ITRN1(I40E_TX_ITR, v_idx - 1), q_vector->tx.itr);
+	wr32(hw, I40E_VFINT_ITRN1(I40E_TX_ITR, q_vector->reg_idx),
+	     q_vector->tx.itr);
 }
 
 /**
-- 
2.14.3

[net-next 08/12] i40e/i40evf: Don't bother setting the CLEARPBA bit

[Jeff Kirsher]

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

The CLEARPBA bit in the dynamic interrupt control register actually has
no effect either way on the hardware. As per errata 28 in the XL710
specification update the interrupt is actually cleared any time the
register is written with the INTENA_MSK bit set to 0. As such the act of
toggling the enable bit actually will trigger the interrupt being
cleared and could lead to potential lost events if auto-masking is
not enabled.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e_txrx.c   | 11 ++++++++++-
 drivers/net/ethernet/intel/i40evf/i40e_txrx.c | 11 ++++++++++-
 2 files changed, 20 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
index 6b3fe5d26dc1..ac1fa9e3c04f 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
@@ -2281,8 +2281,17 @@ static u32 i40e_buildreg_itr(const int type, const u16 itr)
 {
 	u32 val;
 
+	/* We don't bother with setting the CLEARPBA bit as the data sheet
+	 * points out doing so is "meaningless since it was already
+	 * auto-cleared". The auto-clearing happens when the interrupt is
+	 * asserted.
+	 *
+	 * Hardware errata 28 for also indicates that writing to a
+	 * xxINT_DYN_CTLx CSR with INTENA_MSK (bit 31) set to 0 will clear
+	 * an event in the PBA anyway so we need to rely on the automask
+	 * to hold pending events for us until the interrupt is re-enabled
+	 */
 	val = I40E_PFINT_DYN_CTLN_INTENA_MASK |
-	      I40E_PFINT_DYN_CTLN_CLEARPBA_MASK |
 	      (type << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT) |
 	      (itr << I40E_PFINT_DYN_CTLN_INTERVAL_SHIFT);
 
diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
index 3fd7e9731f49..34d898f0adaa 100644
--- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
@@ -1464,8 +1464,17 @@ static u32 i40e_buildreg_itr(const int type, const u16 itr)
 {
 	u32 val;
 
+	/* We don't bother with setting the CLEARPBA bit as the data sheet
+	 * points out doing so is "meaningless since it was already
+	 * auto-cleared". The auto-clearing happens when the interrupt is
+	 * asserted.
+	 *
+	 * Hardware errata 28 for also indicates that writing to a
+	 * xxINT_DYN_CTLx CSR with INTENA_MSK (bit 31) set to 0 will clear
+	 * an event in the PBA anyway so we need to rely on the automask
+	 * to hold pending events for us until the interrupt is re-enabled
+	 */
 	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);
 
-- 
2.14.3

[net-next 09/12] i40e/i40evf: Use usec value instead of reg value for ITR defines

[Jeff Kirsher]

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

Instead of using the register value for the defines when setting up the
ring ITR we can just use the actual values and avoid the use of shifts and
macros to translate between the values we have and the values we want.

This helps to make the code more readable as we can quickly translate from
one value to the other.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e_ethtool.c     | 12 +++---
 drivers/net/ethernet/intel/i40e/i40e_txrx.c        | 11 +++++-
 drivers/net/ethernet/intel/i40e/i40e_txrx.h        | 45 ++++++++++++----------
 drivers/net/ethernet/intel/i40evf/i40e_txrx.c      | 11 +++++-
 drivers/net/ethernet/intel/i40evf/i40e_txrx.h      | 44 +++++++++++----------
 drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c | 12 +++---
 6 files changed, 79 insertions(+), 56 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
index 8cc9198ac32f..3647af8fe32e 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
@@ -2315,8 +2315,8 @@ static void i40e_set_itr_per_queue(struct i40e_vsi *vsi,
 
 	intrl = i40e_intrl_usec_to_reg(vsi->int_rate_limit);
 
-	rx_ring->itr_setting = ec->rx_coalesce_usecs;
-	tx_ring->itr_setting = ec->tx_coalesce_usecs;
+	rx_ring->itr_setting = ITR_REG_ALIGN(ec->rx_coalesce_usecs);
+	tx_ring->itr_setting = ITR_REG_ALIGN(ec->tx_coalesce_usecs);
 
 	if (ec->use_adaptive_rx_coalesce)
 		rx_ring->itr_setting |= I40E_ITR_DYNAMIC;
@@ -2396,7 +2396,7 @@ static int __i40e_set_coalesce(struct net_device *netdev,
 		return -EINVAL;
 	}
 
-	if (ec->rx_coalesce_usecs > (I40E_MAX_ITR << 1)) {
+	if (ec->rx_coalesce_usecs > I40E_MAX_ITR) {
 		netif_info(pf, drv, netdev, "Invalid value, rx-usecs range is 0-8160\n");
 		return -EINVAL;
 	}
@@ -2407,16 +2407,16 @@ static int __i40e_set_coalesce(struct net_device *netdev,
 		return -EINVAL;
 	}
 
-	if (ec->tx_coalesce_usecs > (I40E_MAX_ITR << 1)) {
+	if (ec->tx_coalesce_usecs > I40E_MAX_ITR) {
 		netif_info(pf, drv, netdev, "Invalid value, tx-usecs range is 0-8160\n");
 		return -EINVAL;
 	}
 
 	if (ec->use_adaptive_rx_coalesce && !cur_rx_itr)
-		ec->rx_coalesce_usecs = I40E_MIN_ITR << 1;
+		ec->rx_coalesce_usecs = I40E_MIN_ITR;
 
 	if (ec->use_adaptive_tx_coalesce && !cur_tx_itr)
-		ec->tx_coalesce_usecs = I40E_MIN_ITR << 1;
+		ec->tx_coalesce_usecs = I40E_MIN_ITR;
 
 	intrl_reg = i40e_intrl_usec_to_reg(ec->rx_coalesce_usecs_high);
 	vsi->int_rate_limit = INTRL_REG_TO_USEC(intrl_reg);
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
index ac1fa9e3c04f..ade3e17fba6c 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
@@ -2277,7 +2277,7 @@ static int i40e_clean_rx_irq(struct i40e_ring *rx_ring, int budget)
 	return failure ? budget : (int)total_rx_packets;
 }
 
-static u32 i40e_buildreg_itr(const int type, const u16 itr)
+static inline u32 i40e_buildreg_itr(const int type, u16 itr)
 {
 	u32 val;
 
@@ -2290,10 +2290,17 @@ static u32 i40e_buildreg_itr(const int type, const u16 itr)
 	 * xxINT_DYN_CTLx CSR with INTENA_MSK (bit 31) set to 0 will clear
 	 * an event in the PBA anyway so we need to rely on the automask
 	 * to hold pending events for us until the interrupt is re-enabled
+	 *
+	 * The itr value is reported in microseconds, and the register
+	 * value is recorded in 2 microsecond units. For this reason we
+	 * only need to shift by the interval shift - 1 instead of the
+	 * full value.
 	 */
+	itr &= I40E_ITR_MASK;
+
 	val = I40E_PFINT_DYN_CTLN_INTENA_MASK |
 	      (type << I40E_PFINT_DYN_CTLN_ITR_INDX_SHIFT) |
-	      (itr << I40E_PFINT_DYN_CTLN_INTERVAL_SHIFT);
+	      (itr << (I40E_PFINT_DYN_CTLN_INTERVAL_SHIFT - 1));
 
 	return val;
 }
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.h b/drivers/net/ethernet/intel/i40e/i40e_txrx.h
index 1758dd3bf91b..0f8751c2e595 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.h
@@ -30,32 +30,37 @@
 #include <net/xdp.h>
 
 /* Interrupt Throttling and Rate Limiting Goodies */
-
-#define I40E_MAX_ITR               0x0FF0  /* reg uses 2 usec resolution */
-#define I40E_MIN_ITR               0x0001  /* reg uses 2 usec resolution */
-#define I40E_ITR_100K              0x0005
-#define I40E_ITR_50K               0x000A
-#define I40E_ITR_20K               0x0019
-#define I40E_ITR_18K               0x001B
-#define I40E_ITR_8K                0x003E
-#define I40E_ITR_4K                0x007A
-#define I40E_MAX_INTRL             0x3B    /* reg uses 4 usec resolution */
-#define I40E_ITR_RX_DEF            (ITR_REG_TO_USEC(I40E_ITR_20K) | \
-				    I40E_ITR_DYNAMIC)
-#define I40E_ITR_TX_DEF            (ITR_REG_TO_USEC(I40E_ITR_20K) | \
-				    I40E_ITR_DYNAMIC)
-#define I40E_ITR_DYNAMIC           0x8000  /* use top bit as a flag */
-#define I40E_MIN_INT_RATE          250     /* ~= 1000000 / (I40E_MAX_ITR * 2) */
-#define I40E_MAX_INT_RATE          500000  /* == 1000000 / (I40E_MIN_ITR * 2) */
 #define I40E_DEFAULT_IRQ_WORK      256
-#define ITR_TO_REG(setting) ((setting & ~I40E_ITR_DYNAMIC) >> 1)
-#define ITR_IS_DYNAMIC(setting) (!!(setting & I40E_ITR_DYNAMIC))
-#define ITR_REG_TO_USEC(itr_reg) (itr_reg << 1)
+
+/* The datasheet for the X710 and XL710 indicate that the maximum value for
+ * the ITR is 8160usec which is then called out as 0xFF0 with a 2usec
+ * resolution. 8160 is 0x1FE0 when written out in hex. So instead of storing
+ * the register value which is divided by 2 lets use the actual values and
+ * avoid an excessive amount of translation.
+ */
+#define I40E_ITR_DYNAMIC	0x8000	/* use top bit as a flag */
+#define I40E_ITR_MASK		0x1FFE	/* mask for ITR register value */
+#define I40E_MIN_ITR		     2	/* reg uses 2 usec resolution */
+#define I40E_ITR_100K		    10	/* all values below must be even */
+#define I40E_ITR_50K		    20
+#define I40E_ITR_20K		    50
+#define I40E_ITR_18K		    60
+#define I40E_ITR_8K		   122
+#define I40E_MAX_ITR		  8160	/* maximum value as per datasheet */
+#define ITR_TO_REG(setting) ((setting) & ~I40E_ITR_DYNAMIC)
+#define ITR_REG_ALIGN(setting) __ALIGN_MASK(setting, ~I40E_ITR_MASK)
+#define ITR_IS_DYNAMIC(setting) (!!((setting) & I40E_ITR_DYNAMIC))
+
+#define I40E_ITR_RX_DEF		(I40E_ITR_20K | I40E_ITR_DYNAMIC)
+#define I40E_ITR_TX_DEF		(I40E_ITR_20K | I40E_ITR_DYNAMIC)
+
 /* 0x40 is the enable bit for interrupt rate limiting, and must be set if
  * the value of the rate limit is non-zero
  */
 #define INTRL_ENA                  BIT(6)
+#define I40E_MAX_INTRL             0x3B    /* reg uses 4 usec resolution */
 #define INTRL_REG_TO_USEC(intrl) ((intrl & ~INTRL_ENA) << 2)
+
 /**
  * i40e_intrl_usec_to_reg - convert interrupt rate limit to register
  * @intrl: interrupt rate limit to convert
diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
index 34d898f0adaa..c5f8e941f53e 100644
--- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
@@ -1460,7 +1460,7 @@ static int i40e_clean_rx_irq(struct i40e_ring *rx_ring, int budget)
 	return failure ? budget : (int)total_rx_packets;
 }
 
-static u32 i40e_buildreg_itr(const int type, const u16 itr)
+static inline u32 i40e_buildreg_itr(const int type, u16 itr)
 {
 	u32 val;
 
@@ -1473,10 +1473,17 @@ static u32 i40e_buildreg_itr(const int type, const u16 itr)
 	 * xxINT_DYN_CTLx CSR with INTENA_MSK (bit 31) set to 0 will clear
 	 * an event in the PBA anyway so we need to rely on the automask
 	 * to hold pending events for us until the interrupt is re-enabled
+	 *
+	 * The itr value is reported in microseconds, and the register
+	 * value is recorded in 2 microsecond units. For this reason we
+	 * only need to shift by the interval shift - 1 instead of the
+	 * full value.
 	 */
+	itr &= I40E_ITR_MASK;
+
 	val = I40E_VFINT_DYN_CTLN1_INTENA_MASK |
 	      (type << I40E_VFINT_DYN_CTLN1_ITR_INDX_SHIFT) |
-	      (itr << I40E_VFINT_DYN_CTLN1_INTERVAL_SHIFT);
+	      (itr << (I40E_VFINT_DYN_CTLN1_INTERVAL_SHIFT - 1));
 
 	return val;
 }
diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.h b/drivers/net/ethernet/intel/i40evf/i40e_txrx.h
index 038ed0e2acb7..a4f29722ceff 100644
--- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.h
+++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.h
@@ -28,31 +28,35 @@
 #define _I40E_TXRX_H_
 
 /* Interrupt Throttling and Rate Limiting Goodies */
-
-#define I40E_MAX_ITR               0x0FF0  /* reg uses 2 usec resolution */
-#define I40E_MIN_ITR               0x0001  /* reg uses 2 usec resolution */
-#define I40E_ITR_100K              0x0005
-#define I40E_ITR_50K               0x000A
-#define I40E_ITR_20K               0x0019
-#define I40E_ITR_18K               0x001B
-#define I40E_ITR_8K                0x003E
-#define I40E_ITR_4K                0x007A
-#define I40E_MAX_INTRL             0x3B    /* reg uses 4 usec resolution */
-#define I40E_ITR_RX_DEF            (ITR_REG_TO_USEC(I40E_ITR_20K) | \
-				    I40E_ITR_DYNAMIC)
-#define I40E_ITR_TX_DEF            (ITR_REG_TO_USEC(I40E_ITR_20K) | \
-				    I40E_ITR_DYNAMIC)
-#define I40E_ITR_DYNAMIC           0x8000  /* use top bit as a flag */
-#define I40E_MIN_INT_RATE          250     /* ~= 1000000 / (I40E_MAX_ITR * 2) */
-#define I40E_MAX_INT_RATE          500000  /* == 1000000 / (I40E_MIN_ITR * 2) */
 #define I40E_DEFAULT_IRQ_WORK      256
-#define ITR_TO_REG(setting) ((setting & ~I40E_ITR_DYNAMIC) >> 1)
-#define ITR_IS_DYNAMIC(setting) (!!(setting & I40E_ITR_DYNAMIC))
-#define ITR_REG_TO_USEC(itr_reg) (itr_reg << 1)
+
+/* The datasheet for the X710 and XL710 indicate that the maximum value for
+ * the ITR is 8160usec which is then called out as 0xFF0 with a 2usec
+ * resolution. 8160 is 0x1FE0 when written out in hex. So instead of storing
+ * the register value which is divided by 2 lets use the actual values and
+ * avoid an excessive amount of translation.
+ */
+#define I40E_ITR_DYNAMIC	0x8000	/* use top bit as a flag */
+#define I40E_ITR_MASK		0x1FFE	/* mask for ITR register value */
+#define I40E_MIN_ITR		     2	/* reg uses 2 usec resolution */
+#define I40E_ITR_100K		    10	/* all values below must be even */
+#define I40E_ITR_50K		    20
+#define I40E_ITR_20K		    50
+#define I40E_ITR_18K		    60
+#define I40E_ITR_8K		   122
+#define I40E_MAX_ITR		  8160	/* maximum value as per datasheet */
+#define ITR_TO_REG(setting) ((setting) & ~I40E_ITR_DYNAMIC)
+#define ITR_REG_ALIGN(setting) __ALIGN_MASK(setting, ~I40E_ITR_MASK)
+#define ITR_IS_DYNAMIC(setting) (!!((setting) & I40E_ITR_DYNAMIC))
+
+#define I40E_ITR_RX_DEF		(I40E_ITR_20K | I40E_ITR_DYNAMIC)
+#define I40E_ITR_TX_DEF		(I40E_ITR_20K | I40E_ITR_DYNAMIC)
+
 /* 0x40 is the enable bit for interrupt rate limiting, and must be set if
  * the value of the rate limit is non-zero
  */
 #define INTRL_ENA                  BIT(6)
+#define I40E_MAX_INTRL             0x3B    /* reg uses 4 usec resolution */
 #define INTRL_REG_TO_USEC(intrl) ((intrl & ~INTRL_ENA) << 2)
 #define INTRL_USEC_TO_REG(set) ((set) ? ((set) >> 2) | INTRL_ENA : 0)
 #define I40E_INTRL_8K              125     /* 8000 ints/sec */
diff --git a/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c b/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c
index ed5b8ec4d2a2..f5d372576d71 100644
--- a/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c
+++ b/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c
@@ -517,8 +517,8 @@ static void i40evf_set_itr_per_queue(struct i40evf_adapter *adapter,
 	struct i40e_hw *hw = &adapter->hw;
 	struct i40e_q_vector *q_vector;
 
-	rx_ring->itr_setting = ec->rx_coalesce_usecs;
-	tx_ring->itr_setting = ec->tx_coalesce_usecs;
+	rx_ring->itr_setting = ITR_REG_ALIGN(ec->rx_coalesce_usecs);
+	tx_ring->itr_setting = ITR_REG_ALIGN(ec->tx_coalesce_usecs);
 
 	rx_ring->itr_setting |= I40E_ITR_DYNAMIC;
 	if (!ec->use_adaptive_rx_coalesce)
@@ -563,8 +563,8 @@ static int __i40evf_set_coalesce(struct net_device *netdev,
 	if (ec->rx_coalesce_usecs == 0) {
 		if (ec->use_adaptive_rx_coalesce)
 			netif_info(adapter, drv, netdev, "rx-usecs=0, need to disable adaptive-rx for a complete disable\n");
-	} else if ((ec->rx_coalesce_usecs < (I40E_MIN_ITR << 1)) ||
-		   (ec->rx_coalesce_usecs > (I40E_MAX_ITR << 1))) {
+	} else if ((ec->rx_coalesce_usecs < I40E_MIN_ITR) ||
+		   (ec->rx_coalesce_usecs > I40E_MAX_ITR)) {
 		netif_info(adapter, drv, netdev, "Invalid value, rx-usecs range is 0-8160\n");
 		return -EINVAL;
 	}
@@ -573,8 +573,8 @@ static int __i40evf_set_coalesce(struct net_device *netdev,
 	if (ec->tx_coalesce_usecs == 0) {
 		if (ec->use_adaptive_tx_coalesce)
 			netif_info(adapter, drv, netdev, "tx-usecs=0, need to disable adaptive-tx for a complete disable\n");
-	} else if ((ec->tx_coalesce_usecs < (I40E_MIN_ITR << 1)) ||
-		   (ec->tx_coalesce_usecs > (I40E_MAX_ITR << 1))) {
+	} else if ((ec->tx_coalesce_usecs < I40E_MIN_ITR) ||
+		   (ec->tx_coalesce_usecs > I40E_MAX_ITR)) {
 		netif_info(adapter, drv, netdev, "Invalid value, tx-usecs range is 0-8160\n");
 		return -EINVAL;
 	}
-- 
2.14.3

[net-next 10/12] i40evf: Correctly populate rxitr_idx and txitr_idx

[Jeff Kirsher]

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

While testing code for the recent ITR changes I found that updating the Tx
ITR appeared to have no effect with everything defaulting to the Rx ITR. A
bit of digging narrowed it down the fact that we were asking the PF to
associate all causes with ITR 0 as we weren't populating the itr_idx values
for either Rx or Tx.

To correct it I have added the configuration for these values to this
patch. In addition I did some minor clean-up to just add a local pointer
for the vector map instead of dereferencing it based off of the index
repeatedly. In my opinion this makes the resultant code a bit more readable
and saves us a few characters.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 .../net/ethernet/intel/i40evf/i40evf_virtchnl.c    | 24 ++++++++++++++--------
 1 file changed, 15 insertions(+), 9 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40evf/i40evf_virtchnl.c b/drivers/net/ethernet/intel/i40evf/i40evf_virtchnl.c
index 50ce0d6c09ef..d57a67285505 100644
--- a/drivers/net/ethernet/intel/i40evf/i40evf_virtchnl.c
+++ b/drivers/net/ethernet/intel/i40evf/i40evf_virtchnl.c
@@ -344,6 +344,7 @@ void i40evf_disable_queues(struct i40evf_adapter *adapter)
 void i40evf_map_queues(struct i40evf_adapter *adapter)
 {
 	struct virtchnl_irq_map_info *vimi;
+	struct virtchnl_vector_map *vecmap;
 	int v_idx, q_vectors, len;
 	struct i40e_q_vector *q_vector;
 
@@ -367,17 +368,22 @@ void i40evf_map_queues(struct i40evf_adapter *adapter)
 	vimi->num_vectors = adapter->num_msix_vectors;
 	/* Queue vectors first */
 	for (v_idx = 0; v_idx < q_vectors; v_idx++) {
-		q_vector = adapter->q_vectors + v_idx;
-		vimi->vecmap[v_idx].vsi_id = adapter->vsi_res->vsi_id;
-		vimi->vecmap[v_idx].vector_id = v_idx + NONQ_VECS;
-		vimi->vecmap[v_idx].txq_map = q_vector->ring_mask;
-		vimi->vecmap[v_idx].rxq_map = q_vector->ring_mask;
+		q_vector = &adapter->q_vectors[v_idx];
+		vecmap = &vimi->vecmap[v_idx];
+
+		vecmap->vsi_id = adapter->vsi_res->vsi_id;
+		vecmap->vector_id = v_idx + NONQ_VECS;
+		vecmap->txq_map = q_vector->ring_mask;
+		vecmap->rxq_map = q_vector->ring_mask;
+		vecmap->rxitr_idx = I40E_RX_ITR;
+		vecmap->txitr_idx = I40E_TX_ITR;
 	}
 	/* Misc vector last - this is only for AdminQ messages */
-	vimi->vecmap[v_idx].vsi_id = adapter->vsi_res->vsi_id;
-	vimi->vecmap[v_idx].vector_id = 0;
-	vimi->vecmap[v_idx].txq_map = 0;
-	vimi->vecmap[v_idx].rxq_map = 0;
+	vecmap = &vimi->vecmap[v_idx];
+	vecmap->vsi_id = adapter->vsi_res->vsi_id;
+	vecmap->vector_id = 0;
+	vecmap->txq_map = 0;
+	vecmap->rxq_map = 0;
 
 	adapter->aq_required &= ~I40EVF_FLAG_AQ_MAP_VECTORS;
 	i40evf_send_pf_msg(adapter, VIRTCHNL_OP_CONFIG_IRQ_MAP,
-- 
2.14.3

[net-next 11/12] i40e/i40evf: Split container ITR into current_itr and target_itr

[Jeff Kirsher]

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

This patch is mostly prep-work for replacing the current approach to
programming the dynamic aka adaptive ITR. Specifically here what we are
doing is splitting the Tx and Rx ITR each into two separate values.

The first value current_itr represents the current value of the register.

The second value target_itr represents the desired value of the register.

The general plan by doing this is to allow for deferring the update of the
ITR value under certain circumstances. For now we will work with what we
have, but in the future I hope to change the behavior so that we always
only update one ITR at a time using some simple logic to determine which
ITR requires an update.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e_ethtool.c     | 13 +++--
 drivers/net/ethernet/intel/i40e/i40e_main.c        | 22 ++++---
 drivers/net/ethernet/intel/i40e/i40e_txrx.c        | 68 +++++++++++++---------
 drivers/net/ethernet/intel/i40e/i40e_txrx.h        |  3 +-
 drivers/net/ethernet/intel/i40evf/i40e_txrx.c      | 68 +++++++++++++---------
 drivers/net/ethernet/intel/i40evf/i40e_txrx.h      |  3 +-
 drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c | 14 ++---
 drivers/net/ethernet/intel/i40evf/i40evf_main.c    | 10 ++--
 8 files changed, 115 insertions(+), 86 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
index 3647af8fe32e..29a7412b2fa6 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c
@@ -2329,14 +2329,15 @@ static void i40e_set_itr_per_queue(struct i40e_vsi *vsi,
 		tx_ring->itr_setting &= ~I40E_ITR_DYNAMIC;
 
 	q_vector = rx_ring->q_vector;
-	q_vector->rx.itr = ITR_TO_REG(rx_ring->itr_setting);
-	wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, q_vector->reg_idx),
-	     q_vector->rx.itr);
+	q_vector->rx.target_itr = ITR_TO_REG(rx_ring->itr_setting);
 
 	q_vector = tx_ring->q_vector;
-	q_vector->tx.itr = ITR_TO_REG(tx_ring->itr_setting);
-	wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, q_vector->reg_idx),
-	     q_vector->tx.itr);
+	q_vector->tx.target_itr = ITR_TO_REG(tx_ring->itr_setting);
+
+	/* The interrupt handler itself will take care of programming
+	 * the Tx and Rx ITR values based on the values we have entered
+	 * into the q_vector, no need to write the values now.
+	 */
 
 	wr32(hw, I40E_PFINT_RATEN(q_vector->reg_idx), intrl);
 	i40e_flush(hw);
diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c
index a88fdb8bf5f0..39552b3875ee 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_main.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_main.c
@@ -3450,14 +3450,18 @@ static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
 		struct i40e_q_vector *q_vector = vsi->q_vectors[i];
 
 		q_vector->itr_countdown = ITR_COUNTDOWN_START;
-		q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[i]->itr_setting);
+		q_vector->rx.target_itr =
+			ITR_TO_REG(vsi->rx_rings[i]->itr_setting);
 		q_vector->rx.latency_range = I40E_LOW_LATENCY;
 		wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
-		     q_vector->rx.itr);
-		q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[i]->itr_setting);
+		     q_vector->rx.target_itr);
+		q_vector->rx.current_itr = q_vector->rx.target_itr;
+		q_vector->tx.target_itr =
+			ITR_TO_REG(vsi->tx_rings[i]->itr_setting);
 		q_vector->tx.latency_range = I40E_LOW_LATENCY;
 		wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
-		     q_vector->tx.itr);
+		     q_vector->tx.target_itr);
+		q_vector->tx.current_itr = q_vector->tx.target_itr;
 		wr32(hw, I40E_PFINT_RATEN(vector - 1),
 		     i40e_intrl_usec_to_reg(vsi->int_rate_limit));
 
@@ -3559,12 +3563,14 @@ static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
 
 	/* set the ITR configuration */
 	q_vector->itr_countdown = ITR_COUNTDOWN_START;
-	q_vector->rx.itr = ITR_TO_REG(vsi->rx_rings[0]->itr_setting);
+	q_vector->rx.target_itr = ITR_TO_REG(vsi->rx_rings[0]->itr_setting);
 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
-	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.itr);
-	q_vector->tx.itr = ITR_TO_REG(vsi->tx_rings[0]->itr_setting);
+	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.target_itr);
+	q_vector->rx.current_itr = q_vector->rx.target_itr;
+	q_vector->tx.target_itr = ITR_TO_REG(vsi->tx_rings[0]->itr_setting);
 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
-	wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.itr);
+	wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.target_itr);
+	q_vector->tx.current_itr = q_vector->tx.target_itr;
 
 	i40e_enable_misc_int_causes(pf);
 
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
index ade3e17fba6c..f4257b9e6dae 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
@@ -1012,17 +1012,16 @@ void i40e_force_wb(struct i40e_vsi *vsi, struct i40e_q_vector *q_vector)
 static bool i40e_set_new_dynamic_itr(struct i40e_ring_container *rc)
 {
 	enum i40e_latency_range new_latency_range = rc->latency_range;
-	u32 new_itr = rc->itr;
 	int bytes_per_usec;
 	unsigned int usecs, estimated_usecs;
 
 	if (!rc->ring || !ITR_IS_DYNAMIC(rc->ring->itr_setting))
 		return false;
 
-	if (rc->total_packets == 0 || !rc->itr)
+	if (!rc->total_packets || !rc->current_itr)
 		return false;
 
-	usecs = (rc->itr << 1) * ITR_COUNTDOWN_START;
+	usecs = (rc->current_itr << 1) * ITR_COUNTDOWN_START;
 	bytes_per_usec = rc->total_bytes / usecs;
 
 	/* The calculations in this algorithm depend on interrupts actually
@@ -1070,13 +1069,13 @@ static bool i40e_set_new_dynamic_itr(struct i40e_ring_container *rc)
 
 	switch (new_latency_range) {
 	case I40E_LOWEST_LATENCY:
-		new_itr = I40E_ITR_50K;
+		rc->target_itr = I40E_ITR_50K;
 		break;
 	case I40E_LOW_LATENCY:
-		new_itr = I40E_ITR_20K;
+		rc->target_itr = I40E_ITR_20K;
 		break;
 	case I40E_BULK_LATENCY:
-		new_itr = I40E_ITR_18K;
+		rc->target_itr = I40E_ITR_18K;
 		break;
 	default:
 		break;
@@ -1086,11 +1085,7 @@ static bool i40e_set_new_dynamic_itr(struct i40e_ring_container *rc)
 	rc->total_packets = 0;
 	rc->last_itr_update = jiffies;
 
-	if (new_itr != rc->itr) {
-		rc->itr = new_itr;
-		return true;
-	}
-	return false;
+	return rc->target_itr != rc->current_itr;
 }
 
 /**
@@ -2319,7 +2314,7 @@ static inline void i40e_update_enable_itr(struct i40e_vsi *vsi,
 {
 	struct i40e_hw *hw = &vsi->back->hw;
 	bool rx = false, tx = false;
-	u32 txval;
+	u32 intval;
 
 	/* If we don't have MSIX, then we only need to re-enable icr0 */
 	if (!(vsi->back->flags & I40E_FLAG_MSIX_ENABLED)) {
@@ -2327,8 +2322,6 @@ static inline void i40e_update_enable_itr(struct i40e_vsi *vsi,
 		return;
 	}
 
-	txval = i40e_buildreg_itr(I40E_ITR_NONE, 0);
-
 	/* avoid dynamic calculation if in countdown mode */
 	if (q_vector->itr_countdown > 0)
 		goto enable_int;
@@ -2342,26 +2335,43 @@ static inline void i40e_update_enable_itr(struct i40e_vsi *vsi,
 		 * 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);
-		u32 rxval;
-
-		q_vector->tx.itr = q_vector->rx.itr = itr;
-
-		/* 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 = i40e_buildreg_itr(I40E_RX_ITR, itr) | BIT(31);
+		u16 itr = max(q_vector->tx.target_itr,
+			      q_vector->rx.target_itr);
 
-		/* don't check _DOWN because interrupt isn't being enabled */
-		wr32(hw, INTREG(q_vector->reg_idx), rxval);
-
-		txval = i40e_buildreg_itr(I40E_TX_ITR, itr);
+		q_vector->tx.target_itr = itr;
+		q_vector->rx.target_itr = itr;
 	}
 
 enable_int:
+	if (q_vector->rx.target_itr != q_vector->rx.current_itr) {
+		intval = i40e_buildreg_itr(I40E_RX_ITR,
+					   q_vector->rx.target_itr);
+		q_vector->rx.current_itr = q_vector->rx.target_itr;
+
+		if (q_vector->tx.target_itr != q_vector->tx.current_itr) {
+			/* 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)
+			 *
+			 * don't check _DOWN because interrupt isn't being
+			 * enabled
+			 */
+			wr32(hw, INTREG(q_vector->reg_idx),
+			     intval | BIT(31));
+			/* now that Rx is done process Tx update */
+			goto update_tx;
+		}
+	} else if (q_vector->tx.target_itr != q_vector->tx.current_itr) {
+update_tx:
+		intval = i40e_buildreg_itr(I40E_TX_ITR,
+					   q_vector->tx.target_itr);
+		q_vector->tx.current_itr = q_vector->tx.target_itr;
+	} else {
+		intval = i40e_buildreg_itr(I40E_ITR_NONE, 0);
+	}
+
 	if (!test_bit(__I40E_VSI_DOWN, vsi->state))
-		wr32(hw, INTREG(q_vector->reg_idx), txval);
+		wr32(hw, INTREG(q_vector->reg_idx), intval);
 
 	if (q_vector->itr_countdown)
 		q_vector->itr_countdown--;
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.h b/drivers/net/ethernet/intel/i40e/i40e_txrx.h
index 0f8751c2e595..8ad8ffc63579 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.h
@@ -477,7 +477,8 @@ struct i40e_ring_container {
 	unsigned long last_itr_update;	/* jiffies of last ITR update */
 	u16 count;
 	enum i40e_latency_range latency_range;
-	u16 itr;
+	u16 target_itr;			/* target ITR setting for ring(s) */
+	u16 current_itr;		/* current ITR setting for ring(s) */
 };
 
 /* iterator for handling rings in ring container */
diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
index c5f8e941f53e..1f130e931077 100644
--- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
@@ -409,17 +409,16 @@ void i40evf_force_wb(struct i40e_vsi *vsi, struct i40e_q_vector *q_vector)
 static bool i40e_set_new_dynamic_itr(struct i40e_ring_container *rc)
 {
 	enum i40e_latency_range new_latency_range = rc->latency_range;
-	u32 new_itr = rc->itr;
 	int bytes_per_usec;
 	unsigned int usecs, estimated_usecs;
 
 	if (!rc->ring || !ITR_IS_DYNAMIC(rc->ring->itr_setting))
 		return false;
 
-	if (rc->total_packets == 0 || !rc->itr)
+	if (!rc->total_packets || !rc->current_itr)
 		return false;
 
-	usecs = (rc->itr << 1) * ITR_COUNTDOWN_START;
+	usecs = (rc->current_itr << 1) * ITR_COUNTDOWN_START;
 	bytes_per_usec = rc->total_bytes / usecs;
 
 	/* The calculations in this algorithm depend on interrupts actually
@@ -467,13 +466,13 @@ static bool i40e_set_new_dynamic_itr(struct i40e_ring_container *rc)
 
 	switch (new_latency_range) {
 	case I40E_LOWEST_LATENCY:
-		new_itr = I40E_ITR_50K;
+		rc->target_itr = I40E_ITR_50K;
 		break;
 	case I40E_LOW_LATENCY:
-		new_itr = I40E_ITR_20K;
+		rc->target_itr = I40E_ITR_20K;
 		break;
 	case I40E_BULK_LATENCY:
-		new_itr = I40E_ITR_18K;
+		rc->target_itr = I40E_ITR_18K;
 		break;
 	default:
 		break;
@@ -483,11 +482,7 @@ static bool i40e_set_new_dynamic_itr(struct i40e_ring_container *rc)
 	rc->total_packets = 0;
 	rc->last_itr_update = jiffies;
 
-	if (new_itr != rc->itr) {
-		rc->itr = new_itr;
-		return true;
-	}
-	return false;
+	return rc->target_itr != rc->current_itr;
 }
 
 /**
@@ -1502,9 +1497,7 @@ static inline void i40e_update_enable_itr(struct i40e_vsi *vsi,
 {
 	struct i40e_hw *hw = &vsi->back->hw;
 	bool rx = false, tx = false;
-	u32 txval;
-
-	txval = i40e_buildreg_itr(I40E_ITR_NONE, 0);
+	u32 intval;
 
 	/* avoid dynamic calculation if in countdown mode */
 	if (q_vector->itr_countdown > 0)
@@ -1519,26 +1512,43 @@ static inline void i40e_update_enable_itr(struct i40e_vsi *vsi,
 		 * 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);
-		u32 rxval;
-
-		q_vector->tx.itr = q_vector->rx.itr = itr;
-
-		/* 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 = i40e_buildreg_itr(I40E_RX_ITR, itr) | BIT(31);
+		u16 itr = max(q_vector->tx.target_itr,
+			      q_vector->rx.target_itr);
 
-		/* don't check _DOWN because interrupt isn't being enabled */
-		wr32(hw, INTREG(q_vector->reg_idx), rxval);
-
-		txval = i40e_buildreg_itr(I40E_TX_ITR, itr);
+		q_vector->tx.target_itr = itr;
+		q_vector->rx.target_itr = itr;
 	}
 
 enable_int:
+	if (q_vector->rx.target_itr != q_vector->rx.current_itr) {
+		intval = i40e_buildreg_itr(I40E_RX_ITR,
+					   q_vector->rx.target_itr);
+		q_vector->rx.current_itr = q_vector->rx.target_itr;
+
+		if (q_vector->tx.target_itr != q_vector->tx.current_itr) {
+			/* 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)
+			 *
+			 * don't check _DOWN because interrupt isn't being
+			 * enabled
+			 */
+			wr32(hw, INTREG(q_vector->reg_idx),
+			     intval | BIT(31));
+			/* now that Rx is done process Tx update */
+			goto update_tx;
+		}
+	} else if (q_vector->tx.target_itr != q_vector->tx.current_itr) {
+update_tx:
+		intval = i40e_buildreg_itr(I40E_TX_ITR,
+					   q_vector->tx.target_itr);
+		q_vector->tx.current_itr = q_vector->tx.target_itr;
+	} else {
+		intval = i40e_buildreg_itr(I40E_ITR_NONE, 0);
+	}
+
 	if (!test_bit(__I40E_VSI_DOWN, vsi->state))
-		wr32(hw, INTREG(q_vector->reg_idx), txval);
+		wr32(hw, INTREG(q_vector->reg_idx), intval);
 
 	if (q_vector->itr_countdown)
 		q_vector->itr_countdown--;
diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.h b/drivers/net/ethernet/intel/i40evf/i40e_txrx.h
index a4f29722ceff..4069259aef34 100644
--- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.h
+++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.h
@@ -442,7 +442,8 @@ struct i40e_ring_container {
 	unsigned long last_itr_update;	/* jiffies of last ITR update */
 	u16 count;
 	enum i40e_latency_range latency_range;
-	u16 itr;
+	u16 target_itr;			/* target ITR setting for ring(s) */
+	u16 current_itr;		/* current ITR setting for ring(s) */
 };
 
 /* iterator for handling rings in ring container */
diff --git a/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c b/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c
index f5d372576d71..aded3ad7763e 100644
--- a/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c
+++ b/drivers/net/ethernet/intel/i40evf/i40evf_ethtool.c
@@ -514,7 +514,6 @@ static void i40evf_set_itr_per_queue(struct i40evf_adapter *adapter,
 {
 	struct i40e_ring *rx_ring = &adapter->rx_rings[queue];
 	struct i40e_ring *tx_ring = &adapter->tx_rings[queue];
-	struct i40e_hw *hw = &adapter->hw;
 	struct i40e_q_vector *q_vector;
 
 	rx_ring->itr_setting = ITR_REG_ALIGN(ec->rx_coalesce_usecs);
@@ -529,16 +528,15 @@ static void i40evf_set_itr_per_queue(struct i40evf_adapter *adapter,
 		tx_ring->itr_setting ^= I40E_ITR_DYNAMIC;
 
 	q_vector = rx_ring->q_vector;
-	q_vector->rx.itr = ITR_TO_REG(rx_ring->itr_setting);
-	wr32(hw, I40E_VFINT_ITRN1(I40E_RX_ITR, q_vector->reg_idx),
-	     q_vector->rx.itr);
+	q_vector->rx.target_itr = ITR_TO_REG(rx_ring->itr_setting);
 
 	q_vector = tx_ring->q_vector;
-	q_vector->tx.itr = ITR_TO_REG(tx_ring->itr_setting);
-	wr32(hw, I40E_VFINT_ITRN1(I40E_TX_ITR, q_vector->reg_idx),
-	     q_vector->tx.itr);
+	q_vector->tx.target_itr = ITR_TO_REG(tx_ring->itr_setting);
 
-	i40e_flush(hw);
+	/* The interrupt handler itself will take care of programming
+	 * the Tx and Rx ITR values based on the values we have entered
+	 * into the q_vector, no need to write the values now.
+	 */
 }
 
 /**
diff --git a/drivers/net/ethernet/intel/i40evf/i40evf_main.c b/drivers/net/ethernet/intel/i40evf/i40evf_main.c
index f648e5e97529..3bf6a126be72 100644
--- a/drivers/net/ethernet/intel/i40evf/i40evf_main.c
+++ b/drivers/net/ethernet/intel/i40evf/i40evf_main.c
@@ -354,11 +354,12 @@ i40evf_map_vector_to_rxq(struct i40evf_adapter *adapter, int v_idx, int r_idx)
 	q_vector->rx.ring = rx_ring;
 	q_vector->rx.count++;
 	q_vector->rx.latency_range = I40E_LOW_LATENCY;
-	q_vector->rx.itr = ITR_TO_REG(rx_ring->itr_setting);
+	q_vector->rx.target_itr = ITR_TO_REG(rx_ring->itr_setting);
 	q_vector->ring_mask |= BIT(r_idx);
 	q_vector->itr_countdown = ITR_COUNTDOWN_START;
 	wr32(hw, I40E_VFINT_ITRN1(I40E_RX_ITR, q_vector->reg_idx),
-	     q_vector->rx.itr);
+	     q_vector->rx.current_itr);
+	q_vector->rx.current_itr = q_vector->rx.target_itr;
 }
 
 /**
@@ -380,11 +381,12 @@ i40evf_map_vector_to_txq(struct i40evf_adapter *adapter, int v_idx, int t_idx)
 	q_vector->tx.ring = tx_ring;
 	q_vector->tx.count++;
 	q_vector->tx.latency_range = I40E_LOW_LATENCY;
-	q_vector->tx.itr = ITR_TO_REG(tx_ring->itr_setting);
+	q_vector->tx.target_itr = ITR_TO_REG(tx_ring->itr_setting);
 	q_vector->itr_countdown = ITR_COUNTDOWN_START;
 	q_vector->num_ringpairs++;
 	wr32(hw, I40E_VFINT_ITRN1(I40E_TX_ITR, q_vector->reg_idx),
-	     q_vector->tx.itr);
+	     q_vector->tx.target_itr);
+	q_vector->tx.current_itr = q_vector->tx.target_itr;
 }
 
 /**
-- 
2.14.3

[net-next 12/12] i40e/i40evf: Add support for new mechanism of updating adaptive ITR

[Jeff Kirsher]

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

This patch replaces the existing mechanism for determining the correct
value to program for adaptive ITR with yet another new and more
complicated approach.

The basic idea from a 30K foot view is that this new approach will push the
Rx interrupt moderation up so that by default it starts in low latency and
is gradually pushed up into a higher latency setup as long as doing so
increases the number of packets processed, if the number of packets drops
to 4 to 1 per packet we will reset and just base our ITR on the size of the
packets being received. For Tx we leave it floating at a high interrupt
delay and do not pull it down unless we start processing more than 112
packets per interrupt. If we start exceeding that we will cut our interrupt
rates in half until we are back below 112.

The side effect of these patches are that we will be processing more
packets per interrupt. This is both a good and a bad thing as it means we
will not be blocking processing in the case of things like pktgen and XDP,
but we will also be consuming a bit more CPU in the cases of things such as
network throughput tests using netperf.

One delta from this versus the ixgbe version of the changes is that I have
made the interrupt moderation a bit more aggressive when we are in bulk
mode by moving our "goldilocks zone" up from 48 to 96 to 56 to 112. The
main motivation behind moving this is to address the fact that we need to
update less frequently, and have more fine grained control due to the
separate Tx and Rx ITR times.

Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
Tested-by: Andrew Bowers <andrewx.bowers@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
---
 drivers/net/ethernet/intel/i40e/i40e.h          |   3 +-
 drivers/net/ethernet/intel/i40e/i40e_main.c     |  15 +-
 drivers/net/ethernet/intel/i40e/i40e_txrx.c     | 362 ++++++++++++++++--------
 drivers/net/ethernet/intel/i40e/i40e_txrx.h     |  17 +-
 drivers/net/ethernet/intel/i40evf/i40e_txrx.c   | 362 ++++++++++++++++--------
 drivers/net/ethernet/intel/i40evf/i40e_txrx.h   |  17 +-
 drivers/net/ethernet/intel/i40evf/i40evf.h      |   3 +-
 drivers/net/ethernet/intel/i40evf/i40evf_main.c |   6 +-
 8 files changed, 528 insertions(+), 257 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40e/i40e.h b/drivers/net/ethernet/intel/i40e/i40e.h
index 46e9f4e0a02c..ebe795a7f5f9 100644
--- a/drivers/net/ethernet/intel/i40e/i40e.h
+++ b/drivers/net/ethernet/intel/i40e/i40e.h
@@ -824,6 +824,7 @@ struct i40e_q_vector {
 	struct i40e_ring_container rx;
 	struct i40e_ring_container tx;
 
+	u8 itr_countdown;	/* when 0 should adjust adaptive ITR */
 	u8 num_ringpairs;	/* total number of ring pairs in vector */
 
 	cpumask_t affinity_mask;
@@ -832,8 +833,6 @@ struct i40e_q_vector {
 	struct rcu_head rcu;	/* to avoid race with update stats on free */
 	char name[I40E_INT_NAME_STR_LEN];
 	bool arm_wb_state;
-#define ITR_COUNTDOWN_START 100
-	u8 itr_countdown;	/* when 0 should adjust ITR */
 } ____cacheline_internodealigned_in_smp;
 
 /* lan device */
diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c
index 39552b3875ee..70ecd9c3a163 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_main.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_main.c
@@ -3449,19 +3449,20 @@ static void i40e_vsi_configure_msix(struct i40e_vsi *vsi)
 	for (i = 0; i < vsi->num_q_vectors; i++, vector++) {
 		struct i40e_q_vector *q_vector = vsi->q_vectors[i];
 
-		q_vector->itr_countdown = ITR_COUNTDOWN_START;
+		q_vector->rx.next_update = jiffies + 1;
 		q_vector->rx.target_itr =
 			ITR_TO_REG(vsi->rx_rings[i]->itr_setting);
-		q_vector->rx.latency_range = I40E_LOW_LATENCY;
 		wr32(hw, I40E_PFINT_ITRN(I40E_RX_ITR, vector - 1),
 		     q_vector->rx.target_itr);
 		q_vector->rx.current_itr = q_vector->rx.target_itr;
+
+		q_vector->tx.next_update = jiffies + 1;
 		q_vector->tx.target_itr =
 			ITR_TO_REG(vsi->tx_rings[i]->itr_setting);
-		q_vector->tx.latency_range = I40E_LOW_LATENCY;
 		wr32(hw, I40E_PFINT_ITRN(I40E_TX_ITR, vector - 1),
 		     q_vector->tx.target_itr);
 		q_vector->tx.current_itr = q_vector->tx.target_itr;
+
 		wr32(hw, I40E_PFINT_RATEN(vector - 1),
 		     i40e_intrl_usec_to_reg(vsi->int_rate_limit));
 
@@ -3562,13 +3563,12 @@ static void i40e_configure_msi_and_legacy(struct i40e_vsi *vsi)
 	u32 val;
 
 	/* set the ITR configuration */
-	q_vector->itr_countdown = ITR_COUNTDOWN_START;
+	q_vector->rx.next_update = jiffies + 1;
 	q_vector->rx.target_itr = ITR_TO_REG(vsi->rx_rings[0]->itr_setting);
-	q_vector->rx.latency_range = I40E_LOW_LATENCY;
 	wr32(hw, I40E_PFINT_ITR0(I40E_RX_ITR), q_vector->rx.target_itr);
 	q_vector->rx.current_itr = q_vector->rx.target_itr;
+	q_vector->tx.next_update = jiffies + 1;
 	q_vector->tx.target_itr = ITR_TO_REG(vsi->tx_rings[0]->itr_setting);
-	q_vector->tx.latency_range = I40E_LOW_LATENCY;
 	wr32(hw, I40E_PFINT_ITR0(I40E_TX_ITR), q_vector->tx.target_itr);
 	q_vector->tx.current_itr = q_vector->tx.target_itr;
 
@@ -10345,9 +10345,6 @@ static int i40e_vsi_alloc_q_vector(struct i40e_vsi *vsi, int v_idx, int cpu)
 		netif_napi_add(vsi->netdev, &q_vector->napi,
 			       i40e_napi_poll, NAPI_POLL_WEIGHT);
 
-	q_vector->rx.latency_range = I40E_LOW_LATENCY;
-	q_vector->tx.latency_range = I40E_LOW_LATENCY;
-
 	/* tie q_vector and vsi together */
 	vsi->q_vectors[v_idx] = q_vector;
 
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
index f4257b9e6dae..1ec9b1d8023d 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
@@ -995,97 +995,241 @@ void i40e_force_wb(struct i40e_vsi *vsi, struct i40e_q_vector *q_vector)
 	}
 }
 
+static inline bool i40e_container_is_rx(struct i40e_q_vector *q_vector,
+					struct i40e_ring_container *rc)
+{
+	return &q_vector->rx == rc;
+}
+
+static inline unsigned int i40e_itr_divisor(struct i40e_q_vector *q_vector)
+{
+	unsigned int divisor;
+
+	switch (q_vector->vsi->back->hw.phy.link_info.link_speed) {
+	case I40E_LINK_SPEED_40GB:
+		divisor = I40E_ITR_ADAPTIVE_MIN_INC * 1024;
+		break;
+	case I40E_LINK_SPEED_25GB:
+	case I40E_LINK_SPEED_20GB:
+		divisor = I40E_ITR_ADAPTIVE_MIN_INC * 512;
+		break;
+	default:
+	case I40E_LINK_SPEED_10GB:
+		divisor = I40E_ITR_ADAPTIVE_MIN_INC * 256;
+		break;
+	case I40E_LINK_SPEED_1GB:
+	case I40E_LINK_SPEED_100MB:
+		divisor = I40E_ITR_ADAPTIVE_MIN_INC * 32;
+		break;
+	}
+
+	return divisor;
+}
+
 /**
- * i40e_set_new_dynamic_itr - Find new ITR level
+ * i40e_update_itr - update the dynamic ITR value based on statistics
+ * @q_vector: structure containing interrupt and ring information
  * @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
- * pattern.  Constants in this function were computed based on
- * theoretical maximum wire speed and thresholds were set based on
- * testing data as well as attempting to minimize response time
+ * 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 pattern.  Constants in this function were computed
+ * based on theoretical maximum wire speed and thresholds were set based
+ * on testing data as well as attempting to minimize response time
  * while increasing bulk throughput.
  **/
-static bool i40e_set_new_dynamic_itr(struct i40e_ring_container *rc)
+static void i40e_update_itr(struct i40e_q_vector *q_vector,
+			    struct i40e_ring_container *rc)
 {
-	enum i40e_latency_range new_latency_range = rc->latency_range;
-	int bytes_per_usec;
-	unsigned int usecs, estimated_usecs;
+	unsigned int avg_wire_size, packets, bytes, itr;
+	unsigned long next_update = jiffies;
 
+	/* If we don't have any rings just leave ourselves set for maximum
+	 * possible latency so we take ourselves out of the equation.
+	 */
 	if (!rc->ring || !ITR_IS_DYNAMIC(rc->ring->itr_setting))
-		return false;
+		return;
 
-	if (!rc->total_packets || !rc->current_itr)
-		return false;
+	/* For Rx we want to push the delay up and default to low latency.
+	 * for Tx we want to pull the delay down and default to high latency.
+	 */
+	itr = i40e_container_is_rx(q_vector, rc) ?
+	      I40E_ITR_ADAPTIVE_MIN_USECS | I40E_ITR_ADAPTIVE_LATENCY :
+	      I40E_ITR_ADAPTIVE_MAX_USECS | I40E_ITR_ADAPTIVE_LATENCY;
+
+	/* If we didn't update within up to 1 - 2 jiffies we can assume
+	 * that either packets are coming in so slow there hasn't been
+	 * any work, or that there is so much work that NAPI is dealing
+	 * with interrupt moderation and we don't need to do anything.
+	 */
+	if (time_after(next_update, rc->next_update))
+		goto clear_counts;
+
+	/* If itr_countdown is set it means we programmed an ITR within
+	 * the last 4 interrupt cycles. This has a side effect of us
+	 * potentially firing an early interrupt. In order to work around
+	 * this we need to throw out any data received for a few
+	 * interrupts following the update.
+	 */
+	if (q_vector->itr_countdown) {
+		itr = rc->target_itr;
+		goto clear_counts;
+	}
 
-	usecs = (rc->current_itr << 1) * ITR_COUNTDOWN_START;
-	bytes_per_usec = rc->total_bytes / usecs;
+	packets = rc->total_packets;
+	bytes = rc->total_bytes;
 
-	/* The calculations in this algorithm depend on interrupts actually
-	 * firing at the ITR rate. This may not happen if the packet rate is
-	 * really low, or if we've been napi polling. Check to make sure
-	 * that's not the case before we continue.
+	if (i40e_container_is_rx(q_vector, rc)) {
+		/* If Rx there are 1 to 4 packets and bytes are less than
+		 * 9000 assume insufficient data to use bulk rate limiting
+		 * approach unless Tx is already in bulk rate limiting. We
+		 * are likely latency driven.
+		 */
+		if (packets && packets < 4 && bytes < 9000 &&
+		    (q_vector->tx.target_itr & I40E_ITR_ADAPTIVE_LATENCY)) {
+			itr = I40E_ITR_ADAPTIVE_LATENCY;
+			goto adjust_by_size;
+		}
+	} else if (packets < 4) {
+		/* If we have Tx and Rx ITR maxed and Tx ITR is running in
+		 * bulk mode and we are receiving 4 or fewer packets just
+		 * reset the ITR_ADAPTIVE_LATENCY bit for latency mode so
+		 * that the Rx can relax.
+		 */
+		if (rc->target_itr == I40E_ITR_ADAPTIVE_MAX_USECS &&
+		    (q_vector->rx.target_itr & I40E_ITR_MASK) ==
+		     I40E_ITR_ADAPTIVE_MAX_USECS)
+			goto clear_counts;
+	} else if (packets > 32) {
+		/* If we have processed over 32 packets in a single interrupt
+		 * for Tx assume we need to switch over to "bulk" mode.
+		 */
+		rc->target_itr &= ~I40E_ITR_ADAPTIVE_LATENCY;
+	}
+
+	/* We have no packets to actually measure against. This means
+	 * either one of the other queues on this vector is active or
+	 * we are a Tx queue doing TSO with too high of an interrupt rate.
+	 *
+	 * Between 4 and 56 we can assume that our current interrupt delay
+	 * is only slightly too low. As such we should increase it by a small
+	 * fixed amount.
 	 */
-	estimated_usecs = jiffies_to_usecs(jiffies - rc->last_itr_update);
-	if (estimated_usecs > usecs) {
-		new_latency_range = I40E_LOW_LATENCY;
-		goto reset_latency;
+	if (packets < 56) {
+		itr = rc->target_itr + I40E_ITR_ADAPTIVE_MIN_INC;
+		if ((itr & I40E_ITR_MASK) > I40E_ITR_ADAPTIVE_MAX_USECS) {
+			itr &= I40E_ITR_ADAPTIVE_LATENCY;
+			itr += I40E_ITR_ADAPTIVE_MAX_USECS;
+		}
+		goto clear_counts;
 	}
 
-	/* simple throttlerate management
-	 *   0-10MB/s   lowest (50000 ints/s)
-	 *  10-20MB/s   low    (20000 ints/s)
-	 *  20-1249MB/s bulk   (18000 ints/s)
+	if (packets <= 256) {
+		itr = min(q_vector->tx.current_itr, q_vector->rx.current_itr);
+		itr &= I40E_ITR_MASK;
+
+		/* Between 56 and 112 is our "goldilocks" zone where we are
+		 * working out "just right". Just report that our current
+		 * ITR is good for us.
+		 */
+		if (packets <= 112)
+			goto clear_counts;
+
+		/* If packet count is 128 or greater we are likely looking
+		 * at a slight overrun of the delay we want. Try halving
+		 * our delay to see if that will cut the number of packets
+		 * in half per interrupt.
+		 */
+		itr /= 2;
+		itr &= I40E_ITR_MASK;
+		if (itr < I40E_ITR_ADAPTIVE_MIN_USECS)
+			itr = I40E_ITR_ADAPTIVE_MIN_USECS;
+
+		goto clear_counts;
+	}
+
+	/* The paths below assume we are dealing with a bulk ITR since
+	 * number of packets is greater than 256. We are just going to have
+	 * to compute a value and try to bring the count under control,
+	 * though for smaller packet sizes there isn't much we can do as
+	 * NAPI polling will likely be kicking in sooner rather than later.
+	 */
+	itr = I40E_ITR_ADAPTIVE_BULK;
+
+adjust_by_size:
+	/* If packet counts are 256 or greater we can assume we have a gross
+	 * overestimation of what the rate should be. Instead of trying to fine
+	 * tune it just use the formula below to try and dial in an exact value
+	 * give the current packet size of the frame.
+	 */
+	avg_wire_size = bytes / packets;
+
+	/* The following is a crude approximation of:
+	 *  wmem_default / (size + overhead) = desired_pkts_per_int
+	 *  rate / bits_per_byte / (size + ethernet overhead) = pkt_rate
+	 *  (desired_pkt_rate / pkt_rate) * usecs_per_sec = ITR value
 	 *
-	 * 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.
+	 * Assuming wmem_default is 212992 and overhead is 640 bytes per
+	 * packet, (256 skb, 64 headroom, 320 shared info), we can reduce the
+	 * formula down to
+	 *
+	 *  (170 * (size + 24)) / (size + 640) = ITR
+	 *
+	 * We first do some math on the packet size and then finally bitshift
+	 * by 8 after rounding up. We also have to account for PCIe link speed
+	 * difference as ITR scales based on this.
 	 */
-	switch (new_latency_range) {
-	case I40E_LOWEST_LATENCY:
-		if (bytes_per_usec > 10)
-			new_latency_range = I40E_LOW_LATENCY;
-		break;
-	case I40E_LOW_LATENCY:
-		if (bytes_per_usec > 20)
-			new_latency_range = I40E_BULK_LATENCY;
-		else if (bytes_per_usec <= 10)
-			new_latency_range = I40E_LOWEST_LATENCY;
-		break;
-	case I40E_BULK_LATENCY:
-	default:
-		if (bytes_per_usec <= 20)
-			new_latency_range = I40E_LOW_LATENCY;
-		break;
+	if (avg_wire_size <= 60) {
+		/* Start at 250k ints/sec */
+		avg_wire_size = 4096;
+	} else if (avg_wire_size <= 380) {
+		/* 250K ints/sec to 60K ints/sec */
+		avg_wire_size *= 40;
+		avg_wire_size += 1696;
+	} else if (avg_wire_size <= 1084) {
+		/* 60K ints/sec to 36K ints/sec */
+		avg_wire_size *= 15;
+		avg_wire_size += 11452;
+	} else if (avg_wire_size <= 1980) {
+		/* 36K ints/sec to 30K ints/sec */
+		avg_wire_size *= 5;
+		avg_wire_size += 22420;
+	} else {
+		/* plateau at a limit of 30K ints/sec */
+		avg_wire_size = 32256;
 	}
 
-reset_latency:
-	rc->latency_range = new_latency_range;
+	/* If we are in low latency mode halve our delay which doubles the
+	 * rate to somewhere between 100K to 16K ints/sec
+	 */
+	if (itr & I40E_ITR_ADAPTIVE_LATENCY)
+		avg_wire_size /= 2;
 
-	switch (new_latency_range) {
-	case I40E_LOWEST_LATENCY:
-		rc->target_itr = I40E_ITR_50K;
-		break;
-	case I40E_LOW_LATENCY:
-		rc->target_itr = I40E_ITR_20K;
-		break;
-	case I40E_BULK_LATENCY:
-		rc->target_itr = I40E_ITR_18K;
-		break;
-	default:
-		break;
+	/* Resultant value is 256 times larger than it needs to be. This
+	 * gives us room to adjust the value as needed to either increase
+	 * or decrease the value based on link speeds of 10G, 2.5G, 1G, etc.
+	 *
+	 * Use addition as we have already recorded the new latency flag
+	 * for the ITR value.
+	 */
+	itr += DIV_ROUND_UP(avg_wire_size, i40e_itr_divisor(q_vector)) *
+	       I40E_ITR_ADAPTIVE_MIN_INC;
+
+	if ((itr & I40E_ITR_MASK) > I40E_ITR_ADAPTIVE_MAX_USECS) {
+		itr &= I40E_ITR_ADAPTIVE_LATENCY;
+		itr += I40E_ITR_ADAPTIVE_MAX_USECS;
 	}
 
+clear_counts:
+	/* write back value */
+	rc->target_itr = itr;
+
+	/* next update should occur within next jiffy */
+	rc->next_update = next_update + 1;
+
 	rc->total_bytes = 0;
 	rc->total_packets = 0;
-	rc->last_itr_update = jiffies;
-
-	return rc->target_itr != rc->current_itr;
 }
 
 /**
@@ -2303,6 +2447,15 @@ static inline u32 i40e_buildreg_itr(const int type, u16 itr)
 /* a small macro to shorten up some long lines */
 #define INTREG I40E_PFINT_DYN_CTLN
 
+/* The act of updating the ITR will cause it to immediately trigger. In order
+ * to prevent this from throwing off adaptive update statistics we defer the
+ * update so that it can only happen so often. So after either Tx or Rx are
+ * updated we make the adaptive scheme wait until either the ITR completely
+ * expires via the next_update expiration or we have been through at least
+ * 3 interrupts.
+ */
+#define ITR_COUNTDOWN_START 3
+
 /**
  * i40e_update_enable_itr - Update itr and re-enable MSIX interrupt
  * @vsi: the VSI we care about
@@ -2313,7 +2466,6 @@ static inline void i40e_update_enable_itr(struct i40e_vsi *vsi,
 					  struct i40e_q_vector *q_vector)
 {
 	struct i40e_hw *hw = &vsi->back->hw;
-	bool rx = false, tx = false;
 	u32 intval;
 
 	/* If we don't have MSIX, then we only need to re-enable icr0 */
@@ -2322,61 +2474,49 @@ static inline void i40e_update_enable_itr(struct i40e_vsi *vsi,
 		return;
 	}
 
-	/* avoid dynamic calculation if in countdown mode */
-	if (q_vector->itr_countdown > 0)
-		goto enable_int;
+	/* These will do nothing if dynamic updates are not enabled */
+	i40e_update_itr(q_vector, &q_vector->tx);
+	i40e_update_itr(q_vector, &q_vector->rx);
 
-	/* these will return false if dynamic mode is disabled */
-	rx = i40e_set_new_dynamic_itr(&q_vector->rx);
-	tx = i40e_set_new_dynamic_itr(&q_vector->tx);
-
-	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.target_itr,
-			      q_vector->rx.target_itr);
-
-		q_vector->tx.target_itr = itr;
-		q_vector->rx.target_itr = itr;
-	}
-
-enable_int:
-	if (q_vector->rx.target_itr != q_vector->rx.current_itr) {
+	/* This block of logic allows us to get away with only updating
+	 * one ITR value with each interrupt. The idea is to perform a
+	 * pseudo-lazy update with the following criteria.
+	 *
+	 * 1. Rx is given higher priority than Tx if both are in same state
+	 * 2. If we must reduce an ITR that is given highest priority.
+	 * 3. We then give priority to increasing ITR based on amount.
+	 */
+	if (q_vector->rx.target_itr < q_vector->rx.current_itr) {
+		/* Rx ITR needs to be reduced, this is highest priority */
 		intval = i40e_buildreg_itr(I40E_RX_ITR,
 					   q_vector->rx.target_itr);
 		q_vector->rx.current_itr = q_vector->rx.target_itr;
-
-		if (q_vector->tx.target_itr != q_vector->tx.current_itr) {
-			/* 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)
-			 *
-			 * don't check _DOWN because interrupt isn't being
-			 * enabled
-			 */
-			wr32(hw, INTREG(q_vector->reg_idx),
-			     intval | BIT(31));
-			/* now that Rx is done process Tx update */
-			goto update_tx;
-		}
-	} else if (q_vector->tx.target_itr != q_vector->tx.current_itr) {
-update_tx:
+		q_vector->itr_countdown = ITR_COUNTDOWN_START;
+	} else if ((q_vector->tx.target_itr < q_vector->tx.current_itr) ||
+		   ((q_vector->rx.target_itr - q_vector->rx.current_itr) <
+		    (q_vector->tx.target_itr - q_vector->tx.current_itr))) {
+		/* Tx ITR needs to be reduced, this is second priority
+		 * Tx ITR needs to be increased more than Rx, fourth priority
+		 */
 		intval = i40e_buildreg_itr(I40E_TX_ITR,
 					   q_vector->tx.target_itr);
 		q_vector->tx.current_itr = q_vector->tx.target_itr;
+		q_vector->itr_countdown = ITR_COUNTDOWN_START;
+	} else if (q_vector->rx.current_itr != q_vector->rx.target_itr) {
+		/* Rx ITR needs to be increased, third priority */
+		intval = i40e_buildreg_itr(I40E_RX_ITR,
+					   q_vector->rx.target_itr);
+		q_vector->rx.current_itr = q_vector->rx.target_itr;
+		q_vector->itr_countdown = ITR_COUNTDOWN_START;
 	} else {
+		/* No ITR update, lowest priority */
 		intval = i40e_buildreg_itr(I40E_ITR_NONE, 0);
+		if (q_vector->itr_countdown)
+			q_vector->itr_countdown--;
 	}
 
 	if (!test_bit(__I40E_VSI_DOWN, vsi->state))
 		wr32(hw, INTREG(q_vector->reg_idx), intval);
-
-	if (q_vector->itr_countdown)
-		q_vector->itr_countdown--;
-	else
-		q_vector->itr_countdown = ITR_COUNTDOWN_START;
 }
 
 /**
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.h b/drivers/net/ethernet/intel/i40e/i40e_txrx.h
index 8ad8ffc63579..f75a8fe68fcf 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.h
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.h
@@ -463,20 +463,19 @@ static inline void set_ring_xdp(struct i40e_ring *ring)
 	ring->flags |= I40E_TXR_FLAGS_XDP;
 }
 
-enum i40e_latency_range {
-	I40E_LOWEST_LATENCY = 0,
-	I40E_LOW_LATENCY = 1,
-	I40E_BULK_LATENCY = 2,
-};
+#define I40E_ITR_ADAPTIVE_MIN_INC	0x0002
+#define I40E_ITR_ADAPTIVE_MIN_USECS	0x0002
+#define I40E_ITR_ADAPTIVE_MAX_USECS	0x007e
+#define I40E_ITR_ADAPTIVE_LATENCY	0x8000
+#define I40E_ITR_ADAPTIVE_BULK		0x0000
+#define ITR_IS_BULK(x) (!((x) & I40E_ITR_ADAPTIVE_LATENCY))
 
 struct i40e_ring_container {
-	/* array of pointers to rings */
-	struct i40e_ring *ring;
+	struct i40e_ring *ring;		/* pointer to linked list of ring(s) */
+	unsigned long next_update;	/* jiffies value of next update */
 	unsigned int total_bytes;	/* total bytes processed this int */
 	unsigned int total_packets;	/* total packets processed this int */
-	unsigned long last_itr_update;	/* jiffies of last ITR update */
 	u16 count;
-	enum i40e_latency_range latency_range;
 	u16 target_itr;			/* target ITR setting for ring(s) */
 	u16 current_itr;		/* current ITR setting for ring(s) */
 };
diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
index 1f130e931077..eb8f3e327f6b 100644
--- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c
@@ -392,97 +392,241 @@ void i40evf_force_wb(struct i40e_vsi *vsi, struct i40e_q_vector *q_vector)
 	     val);
 }
 
+static inline bool i40e_container_is_rx(struct i40e_q_vector *q_vector,
+					struct i40e_ring_container *rc)
+{
+	return &q_vector->rx == rc;
+}
+
+static inline unsigned int i40e_itr_divisor(struct i40e_q_vector *q_vector)
+{
+	unsigned int divisor;
+
+	switch (q_vector->adapter->link_speed) {
+	case I40E_LINK_SPEED_40GB:
+		divisor = I40E_ITR_ADAPTIVE_MIN_INC * 1024;
+		break;
+	case I40E_LINK_SPEED_25GB:
+	case I40E_LINK_SPEED_20GB:
+		divisor = I40E_ITR_ADAPTIVE_MIN_INC * 512;
+		break;
+	default:
+	case I40E_LINK_SPEED_10GB:
+		divisor = I40E_ITR_ADAPTIVE_MIN_INC * 256;
+		break;
+	case I40E_LINK_SPEED_1GB:
+	case I40E_LINK_SPEED_100MB:
+		divisor = I40E_ITR_ADAPTIVE_MIN_INC * 32;
+		break;
+	}
+
+	return divisor;
+}
+
 /**
- * i40e_set_new_dynamic_itr - Find new ITR level
+ * i40e_update_itr - update the dynamic ITR value based on statistics
+ * @q_vector: structure containing interrupt and ring information
  * @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
- * pattern.  Constants in this function were computed based on
- * theoretical maximum wire speed and thresholds were set based on
- * testing data as well as attempting to minimize response time
+ * 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 pattern.  Constants in this function were computed
+ * based on theoretical maximum wire speed and thresholds were set based
+ * on testing data as well as attempting to minimize response time
  * while increasing bulk throughput.
  **/
-static bool i40e_set_new_dynamic_itr(struct i40e_ring_container *rc)
+static void i40e_update_itr(struct i40e_q_vector *q_vector,
+			    struct i40e_ring_container *rc)
 {
-	enum i40e_latency_range new_latency_range = rc->latency_range;
-	int bytes_per_usec;
-	unsigned int usecs, estimated_usecs;
+	unsigned int avg_wire_size, packets, bytes, itr;
+	unsigned long next_update = jiffies;
 
+	/* If we don't have any rings just leave ourselves set for maximum
+	 * possible latency so we take ourselves out of the equation.
+	 */
 	if (!rc->ring || !ITR_IS_DYNAMIC(rc->ring->itr_setting))
-		return false;
+		return;
 
-	if (!rc->total_packets || !rc->current_itr)
-		return false;
+	/* For Rx we want to push the delay up and default to low latency.
+	 * for Tx we want to pull the delay down and default to high latency.
+	 */
+	itr = i40e_container_is_rx(q_vector, rc) ?
+	      I40E_ITR_ADAPTIVE_MIN_USECS | I40E_ITR_ADAPTIVE_LATENCY :
+	      I40E_ITR_ADAPTIVE_MAX_USECS | I40E_ITR_ADAPTIVE_LATENCY;
+
+	/* If we didn't update within up to 1 - 2 jiffies we can assume
+	 * that either packets are coming in so slow there hasn't been
+	 * any work, or that there is so much work that NAPI is dealing
+	 * with interrupt moderation and we don't need to do anything.
+	 */
+	if (time_after(next_update, rc->next_update))
+		goto clear_counts;
+
+	/* If itr_countdown is set it means we programmed an ITR within
+	 * the last 4 interrupt cycles. This has a side effect of us
+	 * potentially firing an early interrupt. In order to work around
+	 * this we need to throw out any data received for a few
+	 * interrupts following the update.
+	 */
+	if (q_vector->itr_countdown) {
+		itr = rc->target_itr;
+		goto clear_counts;
+	}
 
-	usecs = (rc->current_itr << 1) * ITR_COUNTDOWN_START;
-	bytes_per_usec = rc->total_bytes / usecs;
+	packets = rc->total_packets;
+	bytes = rc->total_bytes;
 
-	/* The calculations in this algorithm depend on interrupts actually
-	 * firing at the ITR rate. This may not happen if the packet rate is
-	 * really low, or if we've been napi polling. Check to make sure
-	 * that's not the case before we continue.
+	if (i40e_container_is_rx(q_vector, rc)) {
+		/* If Rx there are 1 to 4 packets and bytes are less than
+		 * 9000 assume insufficient data to use bulk rate limiting
+		 * approach unless Tx is already in bulk rate limiting. We
+		 * are likely latency driven.
+		 */
+		if (packets && packets < 4 && bytes < 9000 &&
+		    (q_vector->tx.target_itr & I40E_ITR_ADAPTIVE_LATENCY)) {
+			itr = I40E_ITR_ADAPTIVE_LATENCY;
+			goto adjust_by_size;
+		}
+	} else if (packets < 4) {
+		/* If we have Tx and Rx ITR maxed and Tx ITR is running in
+		 * bulk mode and we are receiving 4 or fewer packets just
+		 * reset the ITR_ADAPTIVE_LATENCY bit for latency mode so
+		 * that the Rx can relax.
+		 */
+		if (rc->target_itr == I40E_ITR_ADAPTIVE_MAX_USECS &&
+		    (q_vector->rx.target_itr & I40E_ITR_MASK) ==
+		     I40E_ITR_ADAPTIVE_MAX_USECS)
+			goto clear_counts;
+	} else if (packets > 32) {
+		/* If we have processed over 32 packets in a single interrupt
+		 * for Tx assume we need to switch over to "bulk" mode.
+		 */
+		rc->target_itr &= ~I40E_ITR_ADAPTIVE_LATENCY;
+	}
+
+	/* We have no packets to actually measure against. This means
+	 * either one of the other queues on this vector is active or
+	 * we are a Tx queue doing TSO with too high of an interrupt rate.
+	 *
+	 * Between 4 and 56 we can assume that our current interrupt delay
+	 * is only slightly too low. As such we should increase it by a small
+	 * fixed amount.
 	 */
-	estimated_usecs = jiffies_to_usecs(jiffies - rc->last_itr_update);
-	if (estimated_usecs > usecs) {
-		new_latency_range = I40E_LOW_LATENCY;
-		goto reset_latency;
+	if (packets < 56) {
+		itr = rc->target_itr + I40E_ITR_ADAPTIVE_MIN_INC;
+		if ((itr & I40E_ITR_MASK) > I40E_ITR_ADAPTIVE_MAX_USECS) {
+			itr &= I40E_ITR_ADAPTIVE_LATENCY;
+			itr += I40E_ITR_ADAPTIVE_MAX_USECS;
+		}
+		goto clear_counts;
 	}
 
-	/* simple throttlerate management
-	 *   0-10MB/s   lowest (50000 ints/s)
-	 *  10-20MB/s   low    (20000 ints/s)
-	 *  20-1249MB/s bulk   (18000 ints/s)
+	if (packets <= 256) {
+		itr = min(q_vector->tx.current_itr, q_vector->rx.current_itr);
+		itr &= I40E_ITR_MASK;
+
+		/* Between 56 and 112 is our "goldilocks" zone where we are
+		 * working out "just right". Just report that our current
+		 * ITR is good for us.
+		 */
+		if (packets <= 112)
+			goto clear_counts;
+
+		/* If packet count is 128 or greater we are likely looking
+		 * at a slight overrun of the delay we want. Try halving
+		 * our delay to see if that will cut the number of packets
+		 * in half per interrupt.
+		 */
+		itr /= 2;
+		itr &= I40E_ITR_MASK;
+		if (itr < I40E_ITR_ADAPTIVE_MIN_USECS)
+			itr = I40E_ITR_ADAPTIVE_MIN_USECS;
+
+		goto clear_counts;
+	}
+
+	/* The paths below assume we are dealing with a bulk ITR since
+	 * number of packets is greater than 256. We are just going to have
+	 * to compute a value and try to bring the count under control,
+	 * though for smaller packet sizes there isn't much we can do as
+	 * NAPI polling will likely be kicking in sooner rather than later.
+	 */
+	itr = I40E_ITR_ADAPTIVE_BULK;
+
+adjust_by_size:
+	/* If packet counts are 256 or greater we can assume we have a gross
+	 * overestimation of what the rate should be. Instead of trying to fine
+	 * tune it just use the formula below to try and dial in an exact value
+	 * give the current packet size of the frame.
+	 */
+	avg_wire_size = bytes / packets;
+
+	/* The following is a crude approximation of:
+	 *  wmem_default / (size + overhead) = desired_pkts_per_int
+	 *  rate / bits_per_byte / (size + ethernet overhead) = pkt_rate
+	 *  (desired_pkt_rate / pkt_rate) * usecs_per_sec = ITR value
 	 *
-	 * 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.
+	 * Assuming wmem_default is 212992 and overhead is 640 bytes per
+	 * packet, (256 skb, 64 headroom, 320 shared info), we can reduce the
+	 * formula down to
+	 *
+	 *  (170 * (size + 24)) / (size + 640) = ITR
+	 *
+	 * We first do some math on the packet size and then finally bitshift
+	 * by 8 after rounding up. We also have to account for PCIe link speed
+	 * difference as ITR scales based on this.
 	 */
-	switch (new_latency_range) {
-	case I40E_LOWEST_LATENCY:
-		if (bytes_per_usec > 10)
-			new_latency_range = I40E_LOW_LATENCY;
-		break;
-	case I40E_LOW_LATENCY:
-		if (bytes_per_usec > 20)
-			new_latency_range = I40E_BULK_LATENCY;
-		else if (bytes_per_usec <= 10)
-			new_latency_range = I40E_LOWEST_LATENCY;
-		break;
-	case I40E_BULK_LATENCY:
-	default:
-		if (bytes_per_usec <= 20)
-			new_latency_range = I40E_LOW_LATENCY;
-		break;
+	if (avg_wire_size <= 60) {
+		/* Start at 250k ints/sec */
+		avg_wire_size = 4096;
+	} else if (avg_wire_size <= 380) {
+		/* 250K ints/sec to 60K ints/sec */
+		avg_wire_size *= 40;
+		avg_wire_size += 1696;
+	} else if (avg_wire_size <= 1084) {
+		/* 60K ints/sec to 36K ints/sec */
+		avg_wire_size *= 15;
+		avg_wire_size += 11452;
+	} else if (avg_wire_size <= 1980) {
+		/* 36K ints/sec to 30K ints/sec */
+		avg_wire_size *= 5;
+		avg_wire_size += 22420;
+	} else {
+		/* plateau at a limit of 30K ints/sec */
+		avg_wire_size = 32256;
 	}
 
-reset_latency:
-	rc->latency_range = new_latency_range;
+	/* If we are in low latency mode halve our delay which doubles the
+	 * rate to somewhere between 100K to 16K ints/sec
+	 */
+	if (itr & I40E_ITR_ADAPTIVE_LATENCY)
+		avg_wire_size /= 2;
 
-	switch (new_latency_range) {
-	case I40E_LOWEST_LATENCY:
-		rc->target_itr = I40E_ITR_50K;
-		break;
-	case I40E_LOW_LATENCY:
-		rc->target_itr = I40E_ITR_20K;
-		break;
-	case I40E_BULK_LATENCY:
-		rc->target_itr = I40E_ITR_18K;
-		break;
-	default:
-		break;
+	/* Resultant value is 256 times larger than it needs to be. This
+	 * gives us room to adjust the value as needed to either increase
+	 * or decrease the value based on link speeds of 10G, 2.5G, 1G, etc.
+	 *
+	 * Use addition as we have already recorded the new latency flag
+	 * for the ITR value.
+	 */
+	itr += DIV_ROUND_UP(avg_wire_size, i40e_itr_divisor(q_vector)) *
+	       I40E_ITR_ADAPTIVE_MIN_INC;
+
+	if ((itr & I40E_ITR_MASK) > I40E_ITR_ADAPTIVE_MAX_USECS) {
+		itr &= I40E_ITR_ADAPTIVE_LATENCY;
+		itr += I40E_ITR_ADAPTIVE_MAX_USECS;
 	}
 
+clear_counts:
+	/* write back value */
+	rc->target_itr = itr;
+
+	/* next update should occur within next jiffy */
+	rc->next_update = next_update + 1;
+
 	rc->total_bytes = 0;
 	rc->total_packets = 0;
-	rc->last_itr_update = jiffies;
-
-	return rc->target_itr != rc->current_itr;
 }
 
 /**
@@ -1486,6 +1630,15 @@ static inline u32 i40e_buildreg_itr(const int type, u16 itr)
 /* a small macro to shorten up some long lines */
 #define INTREG I40E_VFINT_DYN_CTLN1
 
+/* The act of updating the ITR will cause it to immediately trigger. In order
+ * to prevent this from throwing off adaptive update statistics we defer the
+ * update so that it can only happen so often. So after either Tx or Rx are
+ * updated we make the adaptive scheme wait until either the ITR completely
+ * expires via the next_update expiration or we have been through at least
+ * 3 interrupts.
+ */
+#define ITR_COUNTDOWN_START 3
+
 /**
  * i40e_update_enable_itr - Update itr and re-enable MSIX interrupt
  * @vsi: the VSI we care about
@@ -1496,64 +1649,51 @@ static inline void i40e_update_enable_itr(struct i40e_vsi *vsi,
 					  struct i40e_q_vector *q_vector)
 {
 	struct i40e_hw *hw = &vsi->back->hw;
-	bool rx = false, tx = false;
 	u32 intval;
 
-	/* avoid dynamic calculation if in countdown mode */
-	if (q_vector->itr_countdown > 0)
-		goto enable_int;
+	/* These will do nothing if dynamic updates are not enabled */
+	i40e_update_itr(q_vector, &q_vector->tx);
+	i40e_update_itr(q_vector, &q_vector->rx);
 
-	/* these will return false if dynamic mode is disabled */
-	rx = i40e_set_new_dynamic_itr(&q_vector->rx);
-	tx = i40e_set_new_dynamic_itr(&q_vector->tx);
-
-	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.target_itr,
-			      q_vector->rx.target_itr);
-
-		q_vector->tx.target_itr = itr;
-		q_vector->rx.target_itr = itr;
-	}
-
-enable_int:
-	if (q_vector->rx.target_itr != q_vector->rx.current_itr) {
+	/* This block of logic allows us to get away with only updating
+	 * one ITR value with each interrupt. The idea is to perform a
+	 * pseudo-lazy update with the following criteria.
+	 *
+	 * 1. Rx is given higher priority than Tx if both are in same state
+	 * 2. If we must reduce an ITR that is given highest priority.
+	 * 3. We then give priority to increasing ITR based on amount.
+	 */
+	if (q_vector->rx.target_itr < q_vector->rx.current_itr) {
+		/* Rx ITR needs to be reduced, this is highest priority */
 		intval = i40e_buildreg_itr(I40E_RX_ITR,
 					   q_vector->rx.target_itr);
 		q_vector->rx.current_itr = q_vector->rx.target_itr;
-
-		if (q_vector->tx.target_itr != q_vector->tx.current_itr) {
-			/* 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)
-			 *
-			 * don't check _DOWN because interrupt isn't being
-			 * enabled
-			 */
-			wr32(hw, INTREG(q_vector->reg_idx),
-			     intval | BIT(31));
-			/* now that Rx is done process Tx update */
-			goto update_tx;
-		}
-	} else if (q_vector->tx.target_itr != q_vector->tx.current_itr) {
-update_tx:
+		q_vector->itr_countdown = ITR_COUNTDOWN_START;
+	} else if ((q_vector->tx.target_itr < q_vector->tx.current_itr) ||
+		   ((q_vector->rx.target_itr - q_vector->rx.current_itr) <
+		    (q_vector->tx.target_itr - q_vector->tx.current_itr))) {
+		/* Tx ITR needs to be reduced, this is second priority
+		 * Tx ITR needs to be increased more than Rx, fourth priority
+		 */
 		intval = i40e_buildreg_itr(I40E_TX_ITR,
 					   q_vector->tx.target_itr);
 		q_vector->tx.current_itr = q_vector->tx.target_itr;
+		q_vector->itr_countdown = ITR_COUNTDOWN_START;
+	} else if (q_vector->rx.current_itr != q_vector->rx.target_itr) {
+		/* Rx ITR needs to be increased, third priority */
+		intval = i40e_buildreg_itr(I40E_RX_ITR,
+					   q_vector->rx.target_itr);
+		q_vector->rx.current_itr = q_vector->rx.target_itr;
+		q_vector->itr_countdown = ITR_COUNTDOWN_START;
 	} else {
+		/* No ITR update, lowest priority */
 		intval = i40e_buildreg_itr(I40E_ITR_NONE, 0);
+		if (q_vector->itr_countdown)
+			q_vector->itr_countdown--;
 	}
 
 	if (!test_bit(__I40E_VSI_DOWN, vsi->state))
 		wr32(hw, INTREG(q_vector->reg_idx), intval);
-
-	if (q_vector->itr_countdown)
-		q_vector->itr_countdown--;
-	else
-		q_vector->itr_countdown = ITR_COUNTDOWN_START;
 }
 
 /**
diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.h b/drivers/net/ethernet/intel/i40evf/i40e_txrx.h
index 4069259aef34..9129447d079b 100644
--- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.h
+++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.h
@@ -428,20 +428,19 @@ static inline void clear_ring_build_skb_enabled(struct i40e_ring *ring)
 	ring->flags &= ~I40E_RXR_FLAGS_BUILD_SKB_ENABLED;
 }
 
-enum i40e_latency_range {
-	I40E_LOWEST_LATENCY = 0,
-	I40E_LOW_LATENCY = 1,
-	I40E_BULK_LATENCY = 2,
-};
+#define I40E_ITR_ADAPTIVE_MIN_INC	0x0002
+#define I40E_ITR_ADAPTIVE_MIN_USECS	0x0002
+#define I40E_ITR_ADAPTIVE_MAX_USECS	0x007e
+#define I40E_ITR_ADAPTIVE_LATENCY	0x8000
+#define I40E_ITR_ADAPTIVE_BULK		0x0000
+#define ITR_IS_BULK(x) (!((x) & I40E_ITR_ADAPTIVE_LATENCY))
 
 struct i40e_ring_container {
-	/* array of pointers to rings */
-	struct i40e_ring *ring;
+	struct i40e_ring *ring;		/* pointer to linked list of ring(s) */
+	unsigned long next_update;	/* jiffies value of next update */
 	unsigned int total_bytes;	/* total bytes processed this int */
 	unsigned int total_packets;	/* total packets processed this int */
-	unsigned long last_itr_update;	/* jiffies of last ITR update */
 	u16 count;
-	enum i40e_latency_range latency_range;
 	u16 target_itr;			/* target ITR setting for ring(s) */
 	u16 current_itr;		/* current ITR setting for ring(s) */
 };
diff --git a/drivers/net/ethernet/intel/i40evf/i40evf.h b/drivers/net/ethernet/intel/i40evf/i40evf.h
index 9690c1ea019e..b6991e8014d8 100644
--- a/drivers/net/ethernet/intel/i40evf/i40evf.h
+++ b/drivers/net/ethernet/intel/i40evf/i40evf.h
@@ -117,9 +117,8 @@ struct i40e_q_vector {
 	struct i40e_ring_container rx;
 	struct i40e_ring_container tx;
 	u32 ring_mask;
+	u8 itr_countdown;	/* when 0 should adjust adaptive ITR */
 	u8 num_ringpairs;	/* total number of ring pairs in vector */
-#define ITR_COUNTDOWN_START 100
-	u8 itr_countdown;	/* when 0 or 1 update ITR */
 	u16 v_idx;		/* index in the vsi->q_vector array. */
 	u16 reg_idx;		/* register index of the interrupt */
 	char name[IFNAMSIZ + 15];
diff --git a/drivers/net/ethernet/intel/i40evf/i40evf_main.c b/drivers/net/ethernet/intel/i40evf/i40evf_main.c
index 3bf6a126be72..6fd09926181a 100644
--- a/drivers/net/ethernet/intel/i40evf/i40evf_main.c
+++ b/drivers/net/ethernet/intel/i40evf/i40evf_main.c
@@ -353,10 +353,9 @@ i40evf_map_vector_to_rxq(struct i40evf_adapter *adapter, int v_idx, int r_idx)
 	rx_ring->vsi = &adapter->vsi;
 	q_vector->rx.ring = rx_ring;
 	q_vector->rx.count++;
-	q_vector->rx.latency_range = I40E_LOW_LATENCY;
+	q_vector->rx.next_update = jiffies + 1;
 	q_vector->rx.target_itr = ITR_TO_REG(rx_ring->itr_setting);
 	q_vector->ring_mask |= BIT(r_idx);
-	q_vector->itr_countdown = ITR_COUNTDOWN_START;
 	wr32(hw, I40E_VFINT_ITRN1(I40E_RX_ITR, q_vector->reg_idx),
 	     q_vector->rx.current_itr);
 	q_vector->rx.current_itr = q_vector->rx.target_itr;
@@ -380,9 +379,8 @@ i40evf_map_vector_to_txq(struct i40evf_adapter *adapter, int v_idx, int t_idx)
 	tx_ring->vsi = &adapter->vsi;
 	q_vector->tx.ring = tx_ring;
 	q_vector->tx.count++;
-	q_vector->tx.latency_range = I40E_LOW_LATENCY;
+	q_vector->tx.next_update = jiffies + 1;
 	q_vector->tx.target_itr = ITR_TO_REG(tx_ring->itr_setting);
-	q_vector->itr_countdown = ITR_COUNTDOWN_START;
 	q_vector->num_ringpairs++;
 	wr32(hw, I40E_VFINT_ITRN1(I40E_TX_ITR, q_vector->reg_idx),
 	     q_vector->tx.target_itr);
-- 
2.14.3

Re: [net-next 00/12][pull request] 40GbE Intel Wired LAN Driver Updates 2018-02-12

[David Miller]

From: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Date: Mon, 12 Feb 2018 12:46:37 -0800

> This series contains updates to i40e and i40evf.
 ...
> The following are changes since commit 9a61df9e5f7471fe5be3e02bd0bed726b2761a54:
>   Merge tag 'kbuild-v4.16-2' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild
> and are available in the git repository at:
>   git://git.kernel.org/pub/scm/linux/kernel/git/jkirsher/next-queue 40GbE

Looks good, pulled, thanks Jeff.