ice: update dim usage and moderation

The driver was having trouble with unreliable latency when doing single
threaded ping-pong tests. This was root caused to the DIM algorithm
landing on a too slow interrupt value, which caused high latency, and it
was especially present when queues were being switched frequently by the
scheduler as happens on default setups today.

In attempting to improve this, we allow the upper rate limit for
interrupts to move to rate limit of 4 microseconds as a max, which means
that no vector can generate more than 250,000 interrupts per second. The
old config was up to 100,000. The driver previously tried to program the
rate limit too frequently and if the receive and transmit side were both
active on the same vector, the INTRL would be set incorrectly, and this
change fixes that issue as a side effect of the redesign.

This driver will operate from now on with a slightly changed DIM table
with more emphasis towards latency sensitivity by having more table
entries with lower latency than with high latency (high being >= 64
microseconds).

The driver also resets the DIM algorithm state with a new stats set when
there is no work done and the data becomes stale (older than 1 second),
for the respective receive or transmit portion of the interrupt.

Add a new helper for setting rate limit, which will be used more
in a followup patch.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Tested-by: Gurucharan G <gurucharanx.g@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>

1 files changed, 49 insertions, 35 deletions

diff --git a/drivers/net/ethernet/intel/ice/ice_txrx.c b/drivers/net/ethernet/intel/ice/ice_txrx.c
index 01ae331927bd..1373b97b117a 100644
--- a/drivers/net/ethernet/intel/ice/ice_txrx.c
+++ b/drivers/net/ethernet/intel/ice/ice_txrx.c
@@ -1259,6 +1259,41 @@ construct_skb:
 	return failure ? budget : (int)total_rx_pkts;
 }
 
+static void __ice_update_sample(struct ice_q_vector *q_vector,
+				struct ice_ring_container *rc,
+				struct dim_sample *sample,
+				bool is_tx)
+{
+	u64 packets = 0, bytes = 0;
+
+	if (is_tx) {
+		struct ice_tx_ring *tx_ring;
+
+		ice_for_each_tx_ring(tx_ring, *rc) {
+			packets += tx_ring->stats.pkts;
+			bytes += tx_ring->stats.bytes;
+		}
+	} else {
+		struct ice_rx_ring *rx_ring;
+
+		ice_for_each_rx_ring(rx_ring, *rc) {
+			packets += rx_ring->stats.pkts;
+			bytes += rx_ring->stats.bytes;
+		}
+	}
+
+	dim_update_sample(q_vector->total_events, packets, bytes, sample);
+	sample->comp_ctr = 0;
+
+	/* if dim settings get stale, like when not updated for 1
+	 * second or longer, force it to start again. This addresses the
+	 * frequent case of an idle queue being switched to by the
+	 * scheduler. The 1,000 here means 1,000 milliseconds.
+	 */
+	if (ktime_ms_delta(sample->time, rc->dim.start_sample.time) >= 1000)
+		rc->dim.state = DIM_START_MEASURE;
+}
+
 /**
  * ice_net_dim - Update net DIM algorithm
  * @q_vector: the vector associated with the interrupt
@@ -1274,34 +1309,16 @@ static void ice_net_dim(struct ice_q_vector *q_vector)
 	struct ice_ring_container *rx = &q_vector->rx;
 
 	if (ITR_IS_DYNAMIC(tx)) {
-		struct dim_sample dim_sample = {};
-		u64 packets = 0, bytes = 0;
-		struct ice_tx_ring *ring;
-
-		ice_for_each_tx_ring(ring, q_vector->tx) {
-			packets += ring->stats.pkts;
-			bytes += ring->stats.bytes;
-		}
-
-		dim_update_sample(q_vector->total_events, packets, bytes,
-				  &dim_sample);
+		struct dim_sample dim_sample;
 
+		__ice_update_sample(q_vector, tx, &dim_sample, true);
 		net_dim(&tx->dim, dim_sample);
 	}
 
 	if (ITR_IS_DYNAMIC(rx)) {
-		struct dim_sample dim_sample = {};
-		u64 packets = 0, bytes = 0;
-		struct ice_rx_ring *ring;
-
-		ice_for_each_rx_ring(ring, q_vector->rx) {
-			packets += ring->stats.pkts;
-			bytes += ring->stats.bytes;
-		}
-
-		dim_update_sample(q_vector->total_events, packets, bytes,
-				  &dim_sample);
+		struct dim_sample dim_sample;
 
+		__ice_update_sample(q_vector, rx, &dim_sample, false);
 		net_dim(&rx->dim, dim_sample);
 	}
 }
@@ -1328,15 +1345,14 @@ static u32 ice_buildreg_itr(u16 itr_idx, u16 itr)
 }
 
 /**
- * ice_update_ena_itr - Update ITR moderation and re-enable MSI-X interrupt
+ * ice_enable_interrupt - re-enable MSI-X interrupt
  * @q_vector: the vector associated with the interrupt to enable
  *
- * Update the net_dim() algorithm and re-enable the interrupt associated with
- * this vector.
- *
- * If the VSI is down, the interrupt will not be re-enabled.
+ * If the VSI is down, the interrupt will not be re-enabled. Also,
+ * when enabling the interrupt always reset the wb_on_itr to false
+ * and trigger a software interrupt to clean out internal state.
  */
-static void ice_update_ena_itr(struct ice_q_vector *q_vector)
+static void ice_enable_interrupt(struct ice_q_vector *q_vector)
 {
 	struct ice_vsi *vsi = q_vector->vsi;
 	bool wb_en = q_vector->wb_on_itr;
@@ -1351,10 +1367,6 @@ static void ice_update_ena_itr(struct ice_q_vector *q_vector)
 	if (wb_en)
 		q_vector->wb_on_itr = false;
 
-	/* This will do nothing if dynamic updates are not enabled. */
-	ice_net_dim(q_vector);
-
-	/* net_dim() updates ITR out-of-band using a work item */
 	itr_val = ice_buildreg_itr(ICE_ITR_NONE, 0);
 	/* trigger an immediate software interrupt when exiting
 	 * busy poll, to make sure to catch any pending cleanups
@@ -1482,10 +1494,12 @@ int ice_napi_poll(struct napi_struct *napi, int 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)))
-		ice_update_ena_itr(q_vector);
-	else
+	if (likely(napi_complete_done(napi, work_done))) {
+		ice_net_dim(q_vector);
+		ice_enable_interrupt(q_vector);
+	} else {
 		ice_set_wb_on_itr(q_vector);
+	}
 
 	return min_t(int, work_done, budget - 1);
 }