[Intel-wired-lan] [net-next v1 2/2] ethernet/intel: consolidate napi and napi exit

Jesse Brandeburg jesse.brandeburg at intel.com
Thu Nov 8 05:40:18 UTC 2018


While reviewing code, I noticed that Eric Dumazet recommends that
drivers check the return code of napi_complete_done, and use that
to decide to enable interrupts or not when exiting poll.  One of
the Intel drivers was already fixed (ixgbe).

Upon looking at the Intel drivers as a whole, we are handling our
polling and napi exit in a few different ways based on whether we
have multiqueue and whether we have tx cleanup included. Several
drivers had the bug of exiting napi with return 0, which appears
to mess up the accounting in the stack.

Consolidate all the napi routines to do best known way of exiting
and to just mostly look like each other.
1) check return code of napi_complete_done to control interrupt enable
2) return the actual amount of work done.
3) return budget immediately if need napi poll again

Tested the changes on e1000e with a high interrupt rate set, and
it shows about an 8% reduction in the CPU utilization when busy
polling because we aren't re-enabling interrupts when we're about
to be polled.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg at intel.com>
---
 drivers/net/ethernet/intel/e100.c             | 10 +++++----
 drivers/net/ethernet/intel/e1000/e1000_main.c | 11 +++++-----
 drivers/net/ethernet/intel/e1000e/netdev.c    | 17 +++++++-------
 drivers/net/ethernet/intel/fm10k/fm10k_main.c | 10 ++++-----
 drivers/net/ethernet/intel/i40e/i40e_txrx.c   |  9 ++++----
 drivers/net/ethernet/intel/iavf/iavf_txrx.c   |  9 ++++----
 drivers/net/ethernet/intel/ice/ice_txrx.c     | 13 ++++++-----
 drivers/net/ethernet/intel/igb/igb_main.c     | 10 +++++----
 drivers/net/ethernet/intel/igbvf/netdev.c     |  9 +++++---
 drivers/net/ethernet/intel/igc/igc_main.c     | 10 +++++----
 .../net/ethernet/intel/ixgbevf/ixgbevf_main.c | 22 +++++++++++--------
 11 files changed, 75 insertions(+), 55 deletions(-)

diff --git a/drivers/net/ethernet/intel/e100.c b/drivers/net/ethernet/intel/e100.c
index 7c4b55482f72..5e5c57db0d3f 100644
--- a/drivers/net/ethernet/intel/e100.c
+++ b/drivers/net/ethernet/intel/e100.c
@@ -2225,11 +2225,13 @@ static int e100_poll(struct napi_struct *napi, int budget)
 	e100_rx_clean(nic, &work_done, budget);
 	e100_tx_clean(nic);
 
-	/* If budget not fully consumed, exit the polling mode */
-	if (work_done < budget) {
-		napi_complete_done(napi, work_done);
+	/* If budget fully consumed, continue polling */
+	if (work_done == budget)
+		return budget;
+
+	/* only re-enable interrupt if stack agrees polling is really done */
+	if (likely(napi_complete_done(napi, work_done)))
 		e100_enable_irq(nic);
-	}
 
 	return work_done;
 }
diff --git a/drivers/net/ethernet/intel/e1000/e1000_main.c b/drivers/net/ethernet/intel/e1000/e1000_main.c
index 43b6d3cec3b3..8fe9af0e2ab7 100644
--- a/drivers/net/ethernet/intel/e1000/e1000_main.c
+++ b/drivers/net/ethernet/intel/e1000/e1000_main.c
@@ -3803,14 +3803,15 @@ static int e1000_clean(struct napi_struct *napi, int budget)
 
 	adapter->clean_rx(adapter, &adapter->rx_ring[0], &work_done, budget);
 
-	if (!tx_clean_complete)
-		work_done = budget;
+	if (!tx_clean_complete || work_done == budget)
+		return budget;
 
-	/* If budget not fully consumed, exit the polling mode */
-	if (work_done < budget) {
+	/* Exit the polling mode, but don't re-enable interrupts if stack might
+	 * poll us due to busy-polling
+	 */
+	if (likely(napi_complete_done(napi, work_done))) {
 		if (likely(adapter->itr_setting & 3))
 			e1000_set_itr(adapter);
-		napi_complete_done(napi, work_done);
 		if (!test_bit(__E1000_DOWN, &adapter->flags))
 			e1000_irq_enable(adapter);
 	}
diff --git a/drivers/net/ethernet/intel/e1000e/netdev.c b/drivers/net/ethernet/intel/e1000e/netdev.c
index 16a73bd9f4cb..698639f50eac 100644
--- a/drivers/net/ethernet/intel/e1000e/netdev.c
+++ b/drivers/net/ethernet/intel/e1000e/netdev.c
@@ -2651,9 +2651,9 @@ static int e1000_alloc_queues(struct e1000_adapter *adapter)
 /**
  * e1000e_poll - NAPI Rx polling callback
  * @napi: struct associated with this polling callback
- * @weight: number of packets driver is allowed to process this poll
+ * @budget: number of packets driver is allowed to process this poll
  **/
-static int e1000e_poll(struct napi_struct *napi, int weight)
+static int e1000e_poll(struct napi_struct *napi, int budget)
 {
 	struct e1000_adapter *adapter = container_of(napi, struct e1000_adapter,
 						     napi);
@@ -2667,16 +2667,17 @@ static int e1000e_poll(struct napi_struct *napi, int weight)
 	    (adapter->rx_ring->ims_val & adapter->tx_ring->ims_val))
 		tx_cleaned = e1000_clean_tx_irq(adapter->tx_ring);
 
-	adapter->clean_rx(adapter->rx_ring, &work_done, weight);
+	adapter->clean_rx(adapter->rx_ring, &work_done, budget);
 
-	if (!tx_cleaned)
-		work_done = weight;
+	if (!tx_cleaned || work_done == budget)
+		return budget;
 
-	/* If weight not fully consumed, exit the polling mode */
-	if (work_done < weight) {
+	/* Exit the polling mode, but don't re-enable interrupts if stack might
+	 * poll us due to busy-polling
+	 */
+	if (likely(napi_complete_done(napi, work_done))) {
 		if (adapter->itr_setting & 3)
 			e1000_set_itr(adapter);
-		napi_complete_done(napi, work_done);
 		if (!test_bit(__E1000_DOWN, &adapter->state)) {
 			if (adapter->msix_entries)
 				ew32(IMS, adapter->rx_ring->ims_val);
diff --git a/drivers/net/ethernet/intel/fm10k/fm10k_main.c b/drivers/net/ethernet/intel/fm10k/fm10k_main.c
index 5b2a50e5798f..6fd15a734324 100644
--- a/drivers/net/ethernet/intel/fm10k/fm10k_main.c
+++ b/drivers/net/ethernet/intel/fm10k/fm10k_main.c
@@ -1465,11 +1465,11 @@ static int fm10k_poll(struct napi_struct *napi, int budget)
 	if (!clean_complete)
 		return budget;
 
-	/* all work done, exit the polling mode */
-	napi_complete_done(napi, work_done);
-
-	/* re-enable the q_vector */
-	fm10k_qv_enable(q_vector);
+	/* Exit the polling mode, but don't re-enable interrupts if stack might
+	 * poll us due to busy-polling
+	 */
+	if (likely(napi_complete_done(napi, work_done)))
+		fm10k_qv_enable(q_vector);
 
 	return min(work_done, budget - 1);
 }
diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
index aef3c89ee79c..4c796ac80daa 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c
@@ -2667,10 +2667,11 @@ int i40e_napi_poll(struct napi_struct *napi, int budget)
 	if (vsi->back->flags & I40E_TXR_FLAGS_WB_ON_ITR)
 		q_vector->arm_wb_state = false;
 
-	/* Work is done so exit the polling mode and re-enable the interrupt */
-	napi_complete_done(napi, work_done);
-
-	i40e_update_enable_itr(vsi, q_vector);
+	/* Exit the polling mode, but don't re-enable interrupts if stack might
+	 * poll us due to busy-polling
+	 */
+	if (likely(napi_complete_done(napi, work_done)))
+		i40e_update_enable_itr(vsi, q_vector);
 
 	return min(work_done, budget - 1);
 }
diff --git a/drivers/net/ethernet/intel/iavf/iavf_txrx.c b/drivers/net/ethernet/intel/iavf/iavf_txrx.c
index fb9bfad96daf..cf965e5eb7ec 100644
--- a/drivers/net/ethernet/intel/iavf/iavf_txrx.c
+++ b/drivers/net/ethernet/intel/iavf/iavf_txrx.c
@@ -1761,10 +1761,11 @@ int iavf_napi_poll(struct napi_struct *napi, int budget)
 	if (vsi->back->flags & IAVF_TXR_FLAGS_WB_ON_ITR)
 		q_vector->arm_wb_state = false;
 
-	/* Work is done so exit the polling mode and re-enable the interrupt */
-	napi_complete_done(napi, work_done);
-
-	iavf_update_enable_itr(vsi, q_vector);
+	/* Exit the polling mode, but don't re-enable interrupts if stack might
+	 * poll us due to busy-polling
+	 */
+	if (likely(napi_complete_done(napi, work_done)))
+		iavf_update_enable_itr(vsi, q_vector);
 
 	return min(work_done, budget - 1);
 }
diff --git a/drivers/net/ethernet/intel/ice/ice_txrx.c b/drivers/net/ethernet/intel/ice/ice_txrx.c
index 5dae968d853e..ffc3e3bbb3cd 100644
--- a/drivers/net/ethernet/intel/ice/ice_txrx.c
+++ b/drivers/net/ethernet/intel/ice/ice_txrx.c
@@ -1103,11 +1103,14 @@ int ice_napi_poll(struct napi_struct *napi, int budget)
 	if (!clean_complete)
 		return budget;
 
-	/* Work is done so exit the polling mode and re-enable the interrupt */
-	napi_complete_done(napi, work_done);
-	if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
-		ice_irq_dynamic_ena(&vsi->back->hw, vsi, q_vector);
-	return 0;
+	/* Exit the polling mode, but don't re-enable interrupts if stack might
+	 * poll us due to busy-polling
+	 */
+	if (likely(napi_complete_done(napi, work_done)))
+		if (test_bit(ICE_FLAG_MSIX_ENA, pf->flags))
+			ice_irq_dynamic_ena(&vsi->back->hw, vsi, q_vector);
+
+	return min(work_done, budget - 1);
 }
 
 /* helper function for building cmd/type/offset */
diff --git a/drivers/net/ethernet/intel/igb/igb_main.c b/drivers/net/ethernet/intel/igb/igb_main.c
index 5df88ad8ac81..f5592910978d 100644
--- a/drivers/net/ethernet/intel/igb/igb_main.c
+++ b/drivers/net/ethernet/intel/igb/igb_main.c
@@ -7753,11 +7753,13 @@ static int igb_poll(struct napi_struct *napi, int budget)
 	if (!clean_complete)
 		return budget;
 
-	/* If not enough Rx work done, exit the polling mode */
-	napi_complete_done(napi, work_done);
-	igb_ring_irq_enable(q_vector);
+	/* Exit the polling mode, but don't re-enable interrupts if stack might
+	 * poll us due to busy-polling
+	 */
+	if (likely(napi_complete_done(napi, work_done)))
+		igb_ring_irq_enable(q_vector);
 
-	return 0;
+	return min(work_done, budget - 1);
 }
 
 /**
diff --git a/drivers/net/ethernet/intel/igbvf/netdev.c b/drivers/net/ethernet/intel/igbvf/netdev.c
index 820d49eb41ab..4eab83faec62 100644
--- a/drivers/net/ethernet/intel/igbvf/netdev.c
+++ b/drivers/net/ethernet/intel/igbvf/netdev.c
@@ -1186,10 +1186,13 @@ static int igbvf_poll(struct napi_struct *napi, int budget)
 
 	igbvf_clean_rx_irq(adapter, &work_done, budget);
 
-	/* If not enough Rx work done, exit the polling mode */
-	if (work_done < budget) {
-		napi_complete_done(napi, work_done);
+	if (work_done == budget)
+		return budget;
 
+	/* Exit the polling mode, but don't re-enable interrupts if stack might
+	 * poll us due to busy-polling
+	 */
+	if (likely(napi_complete_done(napi, work_done))) {
 		if (adapter->requested_itr & 3)
 			igbvf_set_itr(adapter);
 
diff --git a/drivers/net/ethernet/intel/igc/igc_main.c b/drivers/net/ethernet/intel/igc/igc_main.c
index 9d85707e8a81..3ca8d859caf3 100644
--- a/drivers/net/ethernet/intel/igc/igc_main.c
+++ b/drivers/net/ethernet/intel/igc/igc_main.c
@@ -2866,11 +2866,13 @@ static int igc_poll(struct napi_struct *napi, int budget)
 	if (!clean_complete)
 		return budget;
 
-	/* If not enough Rx work done, exit the polling mode */
-	napi_complete_done(napi, work_done);
-	igc_ring_irq_enable(q_vector);
+	/* Exit the polling mode, but don't re-enable interrupts if stack might
+	 * poll us due to busy-polling
+	 */
+	if (likely(napi_complete_done(napi, work_done)))
+		igc_ring_irq_enable(q_vector);
 
-	return 0;
+	return min(work_done, budget - 1);
 }
 
 /**
diff --git a/drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c b/drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c
index 5e47ede7e832..30a3c800a6ea 100644
--- a/drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c
+++ b/drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c
@@ -1293,16 +1293,20 @@ static int ixgbevf_poll(struct napi_struct *napi, int budget)
 	/* If all work not completed, return budget and keep polling */
 	if (!clean_complete)
 		return budget;
-	/* all work done, exit the polling mode */
-	napi_complete_done(napi, work_done);
-	if (adapter->rx_itr_setting == 1)
-		ixgbevf_set_itr(q_vector);
-	if (!test_bit(__IXGBEVF_DOWN, &adapter->state) &&
-	    !test_bit(__IXGBEVF_REMOVING, &adapter->state))
-		ixgbevf_irq_enable_queues(adapter,
-					  BIT(q_vector->v_idx));
 
-	return 0;
+	/* Exit the polling mode, but don't re-enable interrupts if stack might
+	 * poll us due to busy-polling
+	 */
+	if (likely(napi_complete_done(napi, work_done))) {
+		if (adapter->rx_itr_setting == 1)
+			ixgbevf_set_itr(q_vector);
+		if (!test_bit(__IXGBEVF_DOWN, &adapter->state) &&
+		    !test_bit(__IXGBEVF_REMOVING, &adapter->state))
+			ixgbevf_irq_enable_queues(adapter,
+						  BIT(q_vector->v_idx));
+	}
+
+	return min(work_done, budget - 1);
 }
 
 /**
-- 
2.19.1



More information about the Intel-wired-lan mailing list