Thread (3 messages) 3 messages, 2 authors, 2025-08-22

Re: [PATCH] tracing/sched: add 'next_policy' to trace_sched_switch

From: Steven Rostedt <rostedt@goodmis.org>
Date: 2025-08-22 14:02:07
Also in: lkml

On Fri, 22 Aug 2025 18:51:13 +0800
Xiang Gao [off-list ref] wrote:
quoted hunk ↗ jump to hunk
From: gaoxiang17 <redacted>

Sometimes, when analyzing some real-time process issues, it is necessary to know the sched policy.

Show up in the trace as:

      113.457176: sched_switch: prev_comm=kcompactd0 prev_pid=30 prev_prio=120 prev_state=S ==> next_comm=kworker/u4:1 next_pid=27 next_prio=120 next_policy=0
      113.457282: sched_switch: prev_comm=kworker/u4:1 prev_pid=27 prev_prio=120 prev_state=I ==> next_comm=swapper/0 next_pid=0 next_prio=120 next_policy=0
      113.461166: sched_switch: prev_comm=swapper/0 prev_pid=0 prev_prio=120 prev_state=R ==> next_comm=kworker/u4:1 next_pid=27 next_prio=120 next_policy=0

Signed-off-by: gaoxiang17 <redacted>
---
 include/trace/events/sched.h | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/include/trace/events/sched.h b/include/trace/events/sched.h
index 7b2645b50e78..b416b7bafee4 100644
--- a/include/trace/events/sched.h
+++ b/include/trace/events/sched.h
@@ -234,6 +234,7 @@ TRACE_EVENT(sched_switch,
 		__array(	char,	next_comm,	TASK_COMM_LEN	)
 		__field(	pid_t,	next_pid			)
 		__field(	int,	next_prio			)
+		__field(	unsigned int,	next_policy	)
 	),
 
 	TP_fast_assign(
@@ -244,10 +245,11 @@ TRACE_EVENT(sched_switch,
 		memcpy(__entry->next_comm, next->comm, TASK_COMM_LEN);
 		__entry->next_pid	= next->pid;
 		__entry->next_prio	= next->prio;
+		__entry->next_policy	= next->policy;
 		/* XXX SCHED_DEADLINE */
 	),
 
-	TP_printk("prev_comm=%s prev_pid=%d prev_prio=%d prev_state=%s%s ==> next_comm=%s next_pid=%d next_prio=%d",
+	TP_printk("prev_comm=%s prev_pid=%d prev_prio=%d prev_state=%s%s ==> next_comm=%s next_pid=%d next_prio=%d next_policy=%u",
 		__entry->prev_comm, __entry->prev_pid, __entry->prev_prio,
 
 		(__entry->prev_state & (TASK_REPORT_MAX - 1)) ?
@@ -263,7 +265,7 @@ TRACE_EVENT(sched_switch,
 		  "R",
 
 		__entry->prev_state & TASK_REPORT_MAX ? "+" : "",
-		__entry->next_comm, __entry->next_pid, __entry->next_prio)
+		__entry->next_comm, __entry->next_pid, __entry->next_prio, __entry->next_policy)

I'm fine with this change, but I'm not sure how Peter feels about updating
scheduler tracepoints. That said, why not show the policy name?

	TP_printk("prev_comm=%s prev_pid=%d prev_prio=%d prev_state=%s%s ==> next_comm=%s next_pid=%d next_prio=%d next_policy=%s",
	[..]
		__entry->next_comm, __entry->next_pid, __entry->next_prio,
		__print_symbolic(__entry->next_policy,
			{ SCHED_NORMAL,		"normal" },
			{ SCHED_FIFO,		"FIFO" },
			{ SCHED_RR,		"RR" },
			{ SCHED_BATCH,		"batch" },
			{ SCHED_IDLE,		"idle" },
			{ SCHED_DEADLINE,	"deadline" },
			{ SCHED_EXT,		"sched_ext"}))

-- Steve





 );
 
 /*
  
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help