IB/hfi1: Add ring consumer and producers traces

These traces are used to debugging ring issues.

The ipoib_txreq needed to be moved to a header file to allow access from
the trace header file.

The trace changes include:
- new producer/consumer traces
- new allocation deallocation traces
- additional fidelity for SDMA engine prints

Fixes: 4bd00b55c9 ("IB/hfi1: Add AIP tx traces")
Link: https://lore.kernel.org/r/20210913132852.131370.9664.stgit@awfm-01.cornelisnetworks.com
Signed-off-by: Mike Marciniszyn <mike.marciniszyn@cornelisnetworks.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>

drivers/infiniband/hw/hfi1/ipoib.h

@@ -43,6 +43,26 @@ union hfi1_ipoib_flow {
 	} __attribute__((__packed__));
 };
 
+/**
+ * struct ipoib_txreq - IPOIB transmit descriptor
+ * @txreq: sdma transmit request
+ * @sdma_hdr: 9b ib headers
+ * @sdma_status: status returned by sdma engine
+ * @complete: non-zero implies complete
+ * @priv: ipoib netdev private data
+ * @txq: txq on which skb was output
+ * @skb: skb to send
+ */
+struct ipoib_txreq {
+	struct sdma_txreq           txreq;
+	struct hfi1_sdma_header     sdma_hdr;
+	int                         sdma_status;
+	int                         complete;
+	struct hfi1_ipoib_dev_priv *priv;
+	struct hfi1_ipoib_txq      *txq;
+	struct sk_buff             *skb;
+};
+
 /**
  * struct hfi1_ipoib_circ_buf - List of items to be processed
  * @items: ring of items each a power of two size
@@ -56,7 +76,6 @@ union hfi1_ipoib_flow {
  * @complete_txreqs: count of txreqs completed by sdma
  * @head: ring head
  */
-struct ipoib_txreq;
 struct hfi1_ipoib_circ_buf {
 	void *items;
 	u32 max_items;

drivers/infiniband/hw/hfi1/ipoib_tx.c

@@ -22,26 +22,6 @@
 #define CIRC_NEXT(val, size) CIRC_ADD(val, 1, size)
 #define CIRC_PREV(val, size) CIRC_ADD(val, -1, size)
 
-/**
- * struct ipoib_txreq - IPOIB transmit descriptor
- * @txreq: sdma transmit request
- * @sdma_hdr: 9b ib headers
- * @sdma_status: status returned by sdma engine
- * @complete: non-zero implies complete
- * @priv: ipoib netdev private data
- * @txq: txq on which skb was output
- * @skb: skb to send
- */
-struct ipoib_txreq {
-	struct sdma_txreq           txreq;
-	struct hfi1_sdma_header     sdma_hdr;
-	int                         sdma_status;
-	int                         complete;
-	struct hfi1_ipoib_dev_priv *priv;
-	struct hfi1_ipoib_txq      *txq;
-	struct sk_buff             *skb;
-};
-
 struct ipoib_txparms {
 	struct hfi1_devdata        *dd;
 	struct rdma_ah_attr        *ah_attr;
@@ -187,6 +167,7 @@ static int hfi1_ipoib_poll_tx_ring(struct napi_struct *napi, int budget)
 		if (!smp_load_acquire(&tx->complete))
 			break;
 		tx->complete = 0;
+		trace_hfi1_tx_produce(tx, head);
 		hfi1_ipoib_free_tx(tx, budget);
 		head = CIRC_NEXT(head, max_tx);
 		tx =  hfi1_txreq_from_idx(tx_ring, head);
@@ -495,6 +476,7 @@ static int hfi1_ipoib_send_dma_single(struct net_device *dev,
 	}
 
 	tx_ring = &txq->tx_ring;
+	trace_hfi1_tx_consume(tx, tx_ring->tail);
 	/* consume tx */
 	smp_store_release(&tx_ring->tail, CIRC_NEXT(tx_ring->tail, tx_ring->max_items));
 	ret = hfi1_ipoib_submit_tx(txq, tx);
@@ -557,6 +539,7 @@ static int hfi1_ipoib_send_dma_list(struct net_device *dev,
 	}
 
 	tx_ring = &txq->tx_ring;
+	trace_hfi1_tx_consume(tx, tx_ring->tail);
 	/* consume tx */
 	smp_store_release(&tx_ring->tail, CIRC_NEXT(tx_ring->tail, tx_ring->max_items));
 	list_add_tail(&tx->txreq.list, &txq->tx_list);

drivers/infiniband/hw/hfi1/trace_tx.h

@@ -926,11 +926,13 @@ DECLARE_EVENT_CLASS(/* AIP  */
 		 __netif_subqueue_stopped(txq->priv->netdev, txq->q_idx);
 	),
 	TP_printk(/* print  */
-		"[%s] txq %llx idx %u sde %llx head %lx tail %lx flow %x used %u stops %d no_desc %d stopped %u",
+		"[%s] txq %llx idx %u sde %llx:%u cpu %d head %lx tail %lx flow %x used %u stops %d no_desc %d stopped %u",
 		__get_str(dev),
 		(unsigned long long)__entry->txq,
 		__entry->idx,
 		(unsigned long long)__entry->sde,
+		__entry->sde ? __entry->sde->this_idx : 0,
+		__entry->sde ? __entry->sde->cpu : 0,
 		__entry->head,
 		__entry->tail,
 		__entry->flow,
@@ -995,6 +997,65 @@ DEFINE_EVENT(/* xmit_unstopped */
 	TP_ARGS(txq)
 );
 
+DECLARE_EVENT_CLASS(/* AIP  */
+	hfi1_ipoib_tx_template,
+	TP_PROTO(struct ipoib_txreq *tx, u32 idx),
+	TP_ARGS(tx, idx),
+	TP_STRUCT__entry(/* entry */
+		DD_DEV_ENTRY(tx->txq->priv->dd)
+		__field(struct ipoib_txreq *, tx)
+		__field(struct hfi1_ipoib_txq *, txq)
+		__field(struct sk_buff *, skb)
+		__field(ulong, idx)
+	),
+	TP_fast_assign(/* assign */
+		DD_DEV_ASSIGN(tx->txq->priv->dd);
+		__entry->tx = tx;
+		__entry->skb = tx->skb;
+		__entry->txq = tx->txq;
+		__entry->idx = idx;
+	),
+	TP_printk(/* print  */
+		"[%s] tx %llx txq %llx,%u skb %llx idx %lu",
+		__get_str(dev),
+		(unsigned long long)__entry->tx,
+		(unsigned long long)__entry->txq,
+		__entry->txq ? __entry->txq->q_idx : 0,
+		(unsigned long long)__entry->skb,
+		__entry->idx
+	)
+);
+
+DEFINE_EVENT(/* produce */
+	hfi1_ipoib_tx_template, hfi1_tx_produce,
+	TP_PROTO(struct ipoib_txreq *tx, u32 idx),
+	TP_ARGS(tx, idx)
+);
+
+DEFINE_EVENT(/* consume */
+	hfi1_ipoib_tx_template, hfi1_tx_consume,
+	TP_PROTO(struct ipoib_txreq *tx, u32 idx),
+	TP_ARGS(tx, idx)
+);
+
+DEFINE_EVENT(/* alloc_tx */
+	hfi1_ipoib_txq_template, hfi1_txq_alloc_tx,
+	TP_PROTO(struct hfi1_ipoib_txq *txq),
+	TP_ARGS(txq)
+);
+
+DEFINE_EVENT(/* poll */
+	hfi1_ipoib_txq_template, hfi1_txq_poll,
+	TP_PROTO(struct hfi1_ipoib_txq *txq),
+	TP_ARGS(txq)
+);
+
+DEFINE_EVENT(/* complete */
+	hfi1_ipoib_txq_template, hfi1_txq_complete,
+	TP_PROTO(struct hfi1_ipoib_txq *txq),
+	TP_ARGS(txq)
+);
+
 #endif /* __HFI1_TRACE_TX_H */
 
 #undef TRACE_INCLUDE_PATH