RDMA/core: Trace points for diagnosing completion queue issues

Sample trace events:

   kworker/u29:0-300   [007]   120.042217: cq_alloc:             cq.id=4 nr_cqe=161 comp_vector=2 poll_ctx=WORKQUEUE
          <idle>-0     [002]   120.056292: cq_schedule:          cq.id=4
    kworker/2:1H-482   [002]   120.056402: cq_process:           cq.id=4 wake-up took 109 [us] from interrupt
    kworker/2:1H-482   [002]   120.056407: cq_poll:              cq.id=4 requested 16, returned 1
          <idle>-0     [002]   120.067503: cq_schedule:          cq.id=4
    kworker/2:1H-482   [002]   120.067537: cq_process:           cq.id=4 wake-up took 34 [us] from interrupt
    kworker/2:1H-482   [002]   120.067541: cq_poll:              cq.id=4 requested 16, returned 1
          <idle>-0     [002]   120.067657: cq_schedule:          cq.id=4
    kworker/2:1H-482   [002]   120.067672: cq_process:           cq.id=4 wake-up took 15 [us] from interrupt
    kworker/2:1H-482   [002]   120.067674: cq_poll:              cq.id=4 requested 16, returned 1

 ...

         systemd-1     [002]   122.392653: cq_schedule:          cq.id=4
    kworker/2:1H-482   [002]   122.392688: cq_process:           cq.id=4 wake-up took 35 [us] from interrupt
    kworker/2:1H-482   [002]   122.392693: cq_poll:              cq.id=4 requested 16, returned 16
    kworker/2:1H-482   [002]   122.392836: cq_poll:              cq.id=4 requested 16, returned 16
    kworker/2:1H-482   [002]   122.392970: cq_poll:              cq.id=4 requested 16, returned 16
    kworker/2:1H-482   [002]   122.393083: cq_poll:              cq.id=4 requested 16, returned 16
    kworker/2:1H-482   [002]   122.393195: cq_poll:              cq.id=4 requested 16, returned 3

Several features to note in this output:
 - The WCE count and context type are reported at allocation time
 - The CPU and kworker for each CQ is evident
 - The CQ's restracker ID is tagged on each trace event
 - CQ poll scheduling latency is measured
 - Details about how often single completions occur versus multiple
   completions are evident
 - The cost of the ULP's completion handler is recorded

Link: https://lore.kernel.org/r/20191218201815.30584.3481.stgit@manet.1015granger.net
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Parav Pandit <parav@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>

2 files changed, 276 insertions, 0 deletions

diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index 5608e14e3aad..42f28d39f28c 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -1558,6 +1558,11 @@ struct ib_cq {
 	};
 	struct workqueue_struct *comp_wq;
 	struct dim *dim;
+
+	/* updated only by trace points */
+	ktime_t timestamp;
+	bool interrupt;
+
 	/*
 	 * Implementation details of the RDMA core, don't use in drivers:
 	 */
diff --git a/include/trace/events/rdma_core.h b/include/trace/events/rdma_core.h
new file mode 100644
index 000000000000..08f481554e7f
--- /dev/null
+++ b/include/trace/events/rdma_core.h
@@ -0,0 +1,271 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Trace point definitions for core RDMA functions.
+ *
+ * Author: Chuck Lever <chuck.lever@oracle.com>
+ *
+ * Copyright (c) 2019, Oracle and/or its affiliates. All rights reserved.
+ */
+
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM rdma_core
+
+#if !defined(_TRACE_RDMA_CORE_H) || defined(TRACE_HEADER_MULTI_READ)
+#define _TRACE_RDMA_CORE_H
+
+#include <linux/tracepoint.h>
+#include <rdma/ib_verbs.h>
+
+/*
+ * enum ib_poll_context, from include/rdma/ib_verbs.h
+ */
+#define IB_POLL_CTX_LIST			\
+	ib_poll_ctx(DIRECT)			\
+	ib_poll_ctx(SOFTIRQ)			\
+	ib_poll_ctx(WORKQUEUE)			\
+	ib_poll_ctx_end(UNBOUND_WORKQUEUE)
+
+#undef ib_poll_ctx
+#undef ib_poll_ctx_end
+
+#define ib_poll_ctx(x)		TRACE_DEFINE_ENUM(IB_POLL_##x);
+#define ib_poll_ctx_end(x)	TRACE_DEFINE_ENUM(IB_POLL_##x);
+
+IB_POLL_CTX_LIST
+
+#undef ib_poll_ctx
+#undef ib_poll_ctx_end
+
+#define ib_poll_ctx(x)		{ IB_POLL_##x, #x },
+#define ib_poll_ctx_end(x)	{ IB_POLL_##x, #x }
+
+#define rdma_show_ib_poll_ctx(x) \
+		__print_symbolic(x, IB_POLL_CTX_LIST)
+
+/**
+ ** Completion Queue events
+ **/
+
+TRACE_EVENT(cq_schedule,
+	TP_PROTO(
+		struct ib_cq *cq
+	),
+
+	TP_ARGS(cq),
+
+	TP_STRUCT__entry(
+		__field(u32, cq_id)
+	),
+
+	TP_fast_assign(
+		cq->timestamp = ktime_get();
+		cq->interrupt = true;
+
+		__entry->cq_id = cq->res.id;
+	),
+
+	TP_printk("cq.id=%u", __entry->cq_id)
+);
+
+TRACE_EVENT(cq_reschedule,
+	TP_PROTO(
+		struct ib_cq *cq
+	),
+
+	TP_ARGS(cq),
+
+	TP_STRUCT__entry(
+		__field(u32, cq_id)
+	),
+
+	TP_fast_assign(
+		cq->timestamp = ktime_get();
+		cq->interrupt = false;
+
+		__entry->cq_id = cq->res.id;
+	),
+
+	TP_printk("cq.id=%u", __entry->cq_id)
+);
+
+TRACE_EVENT(cq_process,
+	TP_PROTO(
+		const struct ib_cq *cq
+	),
+
+	TP_ARGS(cq),
+
+	TP_STRUCT__entry(
+		__field(u32, cq_id)
+		__field(bool, interrupt)
+		__field(s64, latency)
+	),
+
+	TP_fast_assign(
+		ktime_t latency = ktime_sub(ktime_get(), cq->timestamp);
+
+		__entry->cq_id = cq->res.id;
+		__entry->latency = ktime_to_us(latency);
+		__entry->interrupt = cq->interrupt;
+	),
+
+	TP_printk("cq.id=%u wake-up took %lld [us] from %s",
+		__entry->cq_id, __entry->latency,
+		__entry->interrupt ? "interrupt" : "reschedule"
+	)
+);
+
+TRACE_EVENT(cq_poll,
+	TP_PROTO(
+		const struct ib_cq *cq,
+		int requested,
+		int rc
+	),
+
+	TP_ARGS(cq, requested, rc),
+
+	TP_STRUCT__entry(
+		__field(u32, cq_id)
+		__field(int, requested)
+		__field(int, rc)
+	),
+
+	TP_fast_assign(
+		__entry->cq_id = cq->res.id;
+		__entry->requested = requested;
+		__entry->rc = rc;
+	),
+
+	TP_printk("cq.id=%u requested %d, returned %d",
+		__entry->cq_id, __entry->requested, __entry->rc
+	)
+);
+
+TRACE_EVENT(cq_drain_complete,
+	TP_PROTO(
+		const struct ib_cq *cq
+	),
+
+	TP_ARGS(cq),
+
+	TP_STRUCT__entry(
+		__field(u32, cq_id)
+	),
+
+	TP_fast_assign(
+		__entry->cq_id = cq->res.id;
+	),
+
+	TP_printk("cq.id=%u",
+		__entry->cq_id
+	)
+);
+
+
+TRACE_EVENT(cq_modify,
+	TP_PROTO(
+		const struct ib_cq *cq,
+		u16 comps,
+		u16 usec
+	),
+
+	TP_ARGS(cq, comps, usec),
+
+	TP_STRUCT__entry(
+		__field(u32, cq_id)
+		__field(unsigned int, comps)
+		__field(unsigned int, usec)
+	),
+
+	TP_fast_assign(
+		__entry->cq_id = cq->res.id;
+		__entry->comps = comps;
+		__entry->usec = usec;
+	),
+
+	TP_printk("cq.id=%u comps=%u usec=%u",
+		__entry->cq_id, __entry->comps, __entry->usec
+	)
+);
+
+TRACE_EVENT(cq_alloc,
+	TP_PROTO(
+		const struct ib_cq *cq,
+		int nr_cqe,
+		int comp_vector,
+		enum ib_poll_context poll_ctx
+	),
+
+	TP_ARGS(cq, nr_cqe, comp_vector, poll_ctx),
+
+	TP_STRUCT__entry(
+		__field(u32, cq_id)
+		__field(int, nr_cqe)
+		__field(int, comp_vector)
+		__field(unsigned long, poll_ctx)
+	),
+
+	TP_fast_assign(
+		__entry->cq_id = cq->res.id;
+		__entry->nr_cqe = nr_cqe;
+		__entry->comp_vector = comp_vector;
+		__entry->poll_ctx = poll_ctx;
+	),
+
+	TP_printk("cq.id=%u nr_cqe=%d comp_vector=%d poll_ctx=%s",
+		__entry->cq_id, __entry->nr_cqe, __entry->comp_vector,
+		rdma_show_ib_poll_ctx(__entry->poll_ctx)
+	)
+);
+
+TRACE_EVENT(cq_alloc_error,
+	TP_PROTO(
+		int nr_cqe,
+		int comp_vector,
+		enum ib_poll_context poll_ctx,
+		int rc
+	),
+
+	TP_ARGS(nr_cqe, comp_vector, poll_ctx, rc),
+
+	TP_STRUCT__entry(
+		__field(int, rc)
+		__field(int, nr_cqe)
+		__field(int, comp_vector)
+		__field(unsigned long, poll_ctx)
+	),
+
+	TP_fast_assign(
+		__entry->rc = rc;
+		__entry->nr_cqe = nr_cqe;
+		__entry->comp_vector = comp_vector;
+		__entry->poll_ctx = poll_ctx;
+	),
+
+	TP_printk("nr_cqe=%d comp_vector=%d poll_ctx=%s rc=%d",
+		__entry->nr_cqe, __entry->comp_vector,
+		rdma_show_ib_poll_ctx(__entry->poll_ctx), __entry->rc
+	)
+);
+
+TRACE_EVENT(cq_free,
+	TP_PROTO(
+		const struct ib_cq *cq
+	),
+
+	TP_ARGS(cq),
+
+	TP_STRUCT__entry(
+		__field(u32, cq_id)
+	),
+
+	TP_fast_assign(
+		__entry->cq_id = cq->res.id;
+	),
+
+	TP_printk("cq.id=%u", __entry->cq_id)
+);
+
+#endif /* _TRACE_RDMA_CORE_H */
+
+#include <trace/define_trace.h>