[PATCH v2] sched_ext: Add tracepoint for scheduler exit

sched_ext schedulers have state in BPF programs and kernel. scx_dump
provides kernel state and BPF program state on error, but this is static
in what it can provide.

Add a sched_ext_exit tracepoint in scx_claim_exit() so that BPF programs
can dynamically inspect scheduler specific state at the moment of exit.
Pass the exiting scx_sched so attached programs can read its state, and,
since exits propagate through a hierarchy of sub-schedulers, identify
which scheduler each event belongs to.

Signed-off-by: Pat Somaru <[email protected]>
---
v2: pass sch to the tracepoint and record name/level so exits in a
    scheduler hierarchy are attributable (Tejun)
v1: https://lore.kernel.org/all/[email protected]/

Tested by attaching to the raw tracepoint and unregistering a scheduler:

  # bpftrace -e 'rawtracepoint:sched_ext_exit {
        $sch = (struct scx_sched *)arg0;
        printf("sch=%p name=%s level=%d kind=%ld\n",
               $sch, $sch->ops.name, $sch->level, (int64)arg1); }'

  Attached 1 probe
  sch=0xff1f39c90f1c7000 name=mitosis_1.1.0_x86_64_unknown_linux_gnu level=0 kind=64

 include/trace/events/sched_ext.h | 23 +++++++++++++++++++++++
 kernel/sched/ext/ext.c           |  2 ++
 2 files changed, 25 insertions(+)

diff --git a/include/trace/events/sched_ext.h b/include/trace/events/sched_ext.h
index d1bf5acd59c5..ce1179478e8b 100644
--- a/include/trace/events/sched_ext.h
+++ b/include/trace/events/sched_ext.h
@@ -84,6 +84,29 @@ TRACE_EVENT(sched_ext_bypass_lb,
 	)
 );
 
+TRACE_EVENT(sched_ext_exit,
+
+	TP_PROTO(struct scx_sched *sch, __u32 kind),
+
+	TP_ARGS(sch, kind),
+
+	TP_STRUCT__entry(
+		__string(	name,	sch->ops.name	)
+		__field(	__s32,	level		)
+		__field(	__u32,	kind		)
+	),
+
+	TP_fast_assign(
+		__assign_str(name);
+		__entry->level		= sch->level;
+		__entry->kind		= kind;
+	),
+
+	TP_printk("sched %s level %d kind %u",
+		  __get_str(name), __entry->level, __entry->kind
+	)
+);
+
 #endif /* _TRACE_SCHED_EXT_H */
 
 /* This part must be outside protection */
diff --git a/kernel/sched/ext/ext.c b/kernel/sched/ext/ext.c
index 1a0ec985da77..6ec774435cd1 100644
--- a/kernel/sched/ext/ext.c
+++ b/kernel/sched/ext/ext.c
@@ -5730,6 +5730,8 @@ static bool scx_claim_exit(struct scx_sched *sch, enum scx_exit_kind kind)
 	 */
 	WRITE_ONCE(sch->aborting, true);
 
+	trace_sched_ext_exit(sch, kind);
+
 	/*
 	 * Propagate exits to descendants immediately. Each has a dedicated
 	 * helper kthread and can run in parallel. While most of disabling is

base-commit: 57194a3172ba0123e8f37c4574a8e2863ab67622
-- 
2.55.0