[PATCH v2 36/36] net/ionic: add watchdogs to protect each queue type

[Andrew Boyer <andrew.boyer@amd.com>]

Ring the doorbell again for the following scenarios:
 * No receives posted but Rx queue not empty after deadline
 * No transmits posted but Tx work still pending after deadline
 * Admin queue work still pending after deadline

This will help the queues recover in the extremely rare case that
a doorbell is missed by the FW.

Signed-off-by: Andrew Boyer <andrew.boyer@amd.com>
---
 drivers/net/ionic/ionic_dev.h         |  4 +++
 drivers/net/ionic/ionic_lif.h         |  3 ++
 drivers/net/ionic/ionic_main.c        | 22 ++++++++++++
 drivers/net/ionic/ionic_rxtx.c        |  1 +
 drivers/net/ionic/ionic_rxtx_sg.c     | 50 ++++++++++++++++++++++++++-
 drivers/net/ionic/ionic_rxtx_simple.c | 50 ++++++++++++++++++++++++++-
 6 files changed, 128 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ionic/ionic_dev.h b/drivers/net/ionic/ionic_dev.h
index 5e238e8ab7..b1e74fbd89 100644
--- a/drivers/net/ionic/ionic_dev.h
+++ b/drivers/net/ionic/ionic_dev.h
@@ -26,6 +26,10 @@
 #define IONIC_DEVCMD_CHECK_PERIOD_US	10	/* devcmd status chk period */
 #define IONIC_DEVCMD_RETRY_WAIT_US	20000
 
+#define IONIC_Q_WDOG_MS			10	/* 10ms */
+#define IONIC_Q_WDOG_MAX_MS		5000	/* 5s */
+#define IONIC_ADMINQ_WDOG_MS		500	/* 500ms */
+
 #define IONIC_ALIGN			4096
 
 struct ionic_adapter;
diff --git a/drivers/net/ionic/ionic_lif.h b/drivers/net/ionic/ionic_lif.h
index 00c8add95c..36b3bcc5a9 100644
--- a/drivers/net/ionic/ionic_lif.h
+++ b/drivers/net/ionic/ionic_lif.h
@@ -85,6 +85,8 @@ struct ionic_rx_qcq {
 	struct rte_mempool *mb_pool;
 	uint64_t rearm_data;
 	uint64_t rearm_seg_data;
+	uint64_t last_wdog_cycles;
+	uint64_t wdog_ms;
 	uint16_t frame_size;	/* Based on configured MTU */
 	uint16_t hdr_seg_size;	/* Length of first segment of RX chain */
 	uint16_t seg_size;	/* Length of all subsequent segments */
@@ -103,6 +105,7 @@ struct ionic_tx_qcq {
 	struct ionic_qcq qcq;
 
 	/* cacheline2 */
+	uint64_t last_wdog_cycles;
 	uint16_t num_segs_fw;	/* # segs supported by current FW */
 	uint16_t free_thresh;
 	uint16_t flags;
diff --git a/drivers/net/ionic/ionic_main.c b/drivers/net/ionic/ionic_main.c
index 26e08d06c9..c957d55bf9 100644
--- a/drivers/net/ionic/ionic_main.c
+++ b/drivers/net/ionic/ionic_main.c
@@ -230,10 +230,16 @@ static int
 ionic_adminq_wait_for_completion(struct ionic_lif *lif,
 		struct ionic_admin_ctx *ctx, unsigned long max_wait)
 {
+	struct ionic_queue *q = &lif->adminqcq->qcq.q;
 	unsigned long step_usec = IONIC_DEVCMD_CHECK_PERIOD_US;
+	unsigned long step_deadline;
 	unsigned long max_wait_usec = max_wait * 1000000L;
 	unsigned long elapsed_usec = 0;
 	int budget = 8;
+	uint16_t idx;
+	void **info;
+
+	step_deadline = IONIC_ADMINQ_WDOG_MS * 1000 / step_usec;
 
 	while (ctx->pending_work && elapsed_usec < max_wait_usec) {
 		/*
@@ -245,10 +251,26 @@ ionic_adminq_wait_for_completion(struct ionic_lif *lif,
 		ionic_qcq_service(&lif->adminqcq->qcq, budget,
 				ionic_adminq_service, NULL);
 
+		/*
+		 * Ring the doorbell again if work is pending after deadline.
+		 */
+		if (ctx->pending_work && !step_deadline) {
+			step_deadline = IONIC_ADMINQ_WDOG_MS *
+				1000 / step_usec;
+
+			rte_spinlock_lock(&lif->adminq_lock);
+			idx = Q_NEXT_TO_POST(q, -1);
+			info = IONIC_INFO_PTR(q, idx);
+			if (info[0] == ctx)
+				ionic_q_flush(q);
+			rte_spinlock_unlock(&lif->adminq_lock);
+		}
+
 		rte_spinlock_unlock(&lif->adminq_service_lock);
 
 		rte_delay_us_block(step_usec);
 		elapsed_usec += step_usec;
+		step_deadline--;
 	}
 
 	return (!ctx->pending_work);
diff --git a/drivers/net/ionic/ionic_rxtx.c b/drivers/net/ionic/ionic_rxtx.c
index d216980450..b9e73b4871 100644
--- a/drivers/net/ionic/ionic_rxtx.c
+++ b/drivers/net/ionic/ionic_rxtx.c
@@ -551,6 +551,7 @@ ionic_dev_rx_queue_setup(struct rte_eth_dev *eth_dev,
 	}
 
 	rxq->mb_pool = mp;
+	rxq->wdog_ms = IONIC_Q_WDOG_MS;
 
 	/*
 	 * Note: the interface does not currently support
diff --git a/drivers/net/ionic/ionic_rxtx_sg.c b/drivers/net/ionic/ionic_rxtx_sg.c
index 4e4a18e5d7..2752ba2acd 100644
--- a/drivers/net/ionic/ionic_rxtx_sg.c
+++ b/drivers/net/ionic/ionic_rxtx_sg.c
@@ -167,6 +167,7 @@ ionic_xmit_pkts_sg(void *tx_queue, struct rte_mbuf **tx_pkts,
 	struct rte_mbuf *mbuf;
 	uint32_t bytes_tx = 0;
 	uint16_t nb_avail, nb_tx = 0;
+	uint64_t then, now, hz, delta;
 	int err;
 
 	struct ionic_txq_desc *desc_base = q->base;
@@ -220,8 +221,26 @@ ionic_xmit_pkts_sg(void *tx_queue, struct rte_mbuf **tx_pkts,
 		rte_wmb();
 		ionic_q_flush(q);
 
+		txq->last_wdog_cycles = rte_get_timer_cycles();
+
 		stats->packets += nb_tx;
 		stats->bytes += bytes_tx;
+	} else {
+		/*
+		 * Ring the doorbell again if no work could be posted and work
+		 * is still pending after the deadline.
+		 */
+		if (q->head_idx != q->tail_idx) {
+			then = txq->last_wdog_cycles;
+			now = rte_get_timer_cycles();
+			hz = rte_get_timer_hz();
+			delta = (now - then) * 1000;
+
+			if (delta >= hz * IONIC_Q_WDOG_MS) {
+				ionic_q_flush(q);
+				txq->last_wdog_cycles = now;
+			}
+		}
 	}
 
 	return nb_tx;
@@ -421,6 +440,7 @@ ionic_rxq_service_sg(struct ionic_rx_qcq *rxq, uint32_t work_to_do,
 	struct ionic_rxq_desc *q_desc_base = q->base;
 	struct ionic_rxq_comp *cq_desc, *cq_desc_base = cq->base;
 	uint32_t work_done = 0;
+	uint64_t then, now, hz, delta;
 
 	cq_desc = &cq_desc_base[cq->tail_idx];
 
@@ -453,8 +473,36 @@ ionic_rxq_service_sg(struct ionic_rx_qcq *rxq, uint32_t work_to_do,
 	}
 
 	/* Update the queue indices and ring the doorbell */
-	if (work_done)
+	if (work_done) {
 		ionic_q_flush(q);
+		rxq->last_wdog_cycles = rte_get_timer_cycles();
+		rxq->wdog_ms = IONIC_Q_WDOG_MS;
+	} else {
+		/*
+		 * Ring the doorbell again if no recvs were posted and the
+		 * recv queue is not empty after the deadline.
+		 *
+		 * Exponentially back off the deadline to avoid excessive
+		 * doorbells when the recv queue is idle.
+		 */
+		if (q->head_idx != q->tail_idx) {
+			then = rxq->last_wdog_cycles;
+			now = rte_get_timer_cycles();
+			hz = rte_get_timer_hz();
+			delta = (now - then) * 1000;
+
+			if (delta >= hz * rxq->wdog_ms) {
+				ionic_q_flush(q);
+				rxq->last_wdog_cycles = now;
+
+				delta = 2 * rxq->wdog_ms;
+				if (delta > IONIC_Q_WDOG_MAX_MS)
+					delta = IONIC_Q_WDOG_MAX_MS;
+
+				rxq->wdog_ms = delta;
+			}
+		}
+	}
 }
 
 uint16_t
diff --git a/drivers/net/ionic/ionic_rxtx_simple.c b/drivers/net/ionic/ionic_rxtx_simple.c
index a191fd3ec9..0421fb32b2 100644
--- a/drivers/net/ionic/ionic_rxtx_simple.c
+++ b/drivers/net/ionic/ionic_rxtx_simple.c
@@ -140,6 +140,7 @@ ionic_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
 	struct rte_mbuf *mbuf;
 	uint32_t bytes_tx = 0;
 	uint16_t nb_avail, nb_tx = 0;
+	uint64_t then, now, hz, delta;
 	int err;
 
 	struct ionic_txq_desc *desc_base = q->base;
@@ -193,8 +194,26 @@ ionic_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts,
 		rte_wmb();
 		ionic_q_flush(q);
 
+		txq->last_wdog_cycles = rte_get_timer_cycles();
+
 		stats->packets += nb_tx;
 		stats->bytes += bytes_tx;
+	} else {
+		/*
+		 * Ring the doorbell again if no work could be posted and work
+		 * is still pending after the deadline.
+		 */
+		if (q->head_idx != q->tail_idx) {
+			then = txq->last_wdog_cycles;
+			now = rte_get_timer_cycles();
+			hz = rte_get_timer_hz();
+			delta = (now - then) * 1000;
+
+			if (delta >= hz * IONIC_Q_WDOG_MS) {
+				ionic_q_flush(q);
+				txq->last_wdog_cycles = now;
+			}
+		}
 	}
 
 	return nb_tx;
@@ -342,6 +361,7 @@ ionic_rxq_service(struct ionic_rx_qcq *rxq, uint32_t work_to_do,
 	struct ionic_rxq_desc *q_desc_base = q->base;
 	struct ionic_rxq_comp *cq_desc, *cq_desc_base = cq->base;
 	uint32_t work_done = 0;
+	uint64_t then, now, hz, delta;
 
 	cq_desc = &cq_desc_base[cq->tail_idx];
 
@@ -374,8 +394,36 @@ ionic_rxq_service(struct ionic_rx_qcq *rxq, uint32_t work_to_do,
 	}
 
 	/* Update the queue indices and ring the doorbell */
-	if (work_done)
+	if (work_done) {
 		ionic_q_flush(q);
+		rxq->last_wdog_cycles = rte_get_timer_cycles();
+		rxq->wdog_ms = IONIC_Q_WDOG_MS;
+	} else {
+		/*
+		 * Ring the doorbell again if no recvs were posted and the
+		 * recv queue is not empty after the deadline.
+		 *
+		 * Exponentially back off the deadline to avoid excessive
+		 * doorbells when the recv queue is idle.
+		 */
+		if (q->head_idx != q->tail_idx) {
+			then = rxq->last_wdog_cycles;
+			now = rte_get_timer_cycles();
+			hz = rte_get_timer_hz();
+			delta = (now - then) * 1000;
+
+			if (delta >= hz * rxq->wdog_ms) {
+				ionic_q_flush(q);
+				rxq->last_wdog_cycles = now;
+
+				delta = 2 * rxq->wdog_ms;
+				if (delta > IONIC_Q_WDOG_MAX_MS)
+					delta = IONIC_Q_WDOG_MAX_MS;
+
+				rxq->wdog_ms = delta;
+			}
+		}
+	}
 }
 
 uint16_t
-- 
2.17.1