public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Oleg Nesterov <oleg@redhat.com>
To: Masami Hiramatsu <masami.hiramatsu.pt@hitachi.com>,
	Steven Rostedt <rostedt@goodmis.org>
Cc: Frederic Weisbecker <fweisbec@gmail.com>,
	Ingo Molnar <mingo@redhat.com>,
	Srikar Dronamraju <srikar@linux.vnet.ibm.com>,
	"zhangwei(Jovi)" <jovi.zhangwei@huawei.com>,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH 0/3] tracing: more list_empty(perf_events) checks
Date: Mon, 17 Jun 2013 22:18:18 +0200	[thread overview]
Message-ID: <20130617201818.GA12349@redhat.com> (raw)
In-Reply-To: <20130617170142.GA19780@redhat.com>

On 06/17, Oleg Nesterov wrote:
>
> DECLARE_EVENT_CLASS()->perf_trace_##call() is not trivial because
> of __perf_task()

Perhaps we can do something like below?

Then we can

	1. kill __perf_addr(), __perf_count(), __perf_task() and
	   TP_perf_assign()

	2. Add the fast path check

		if (builtin_constant(__task) && !__task &&
		    hlist_empty(head)))
			return;

	   into perf_trace_call() right after ftrace_get_offsets_call().

Doesn't look very nice, it relies on the fact that
ftrace_get_offsets_call(args) evaluates TP_perf_arg()'s hidden in
TP_ARGS().

OTOH, the whole point of include/trace/ is to create the code which
nobody except the maintainers can understand. At least I certainly
can't ;) So perhaps this is fine.

Oleg.

--- a/include/trace/events/sched.h
+++ b/include/trace/events/sched.h
@@ -57,7 +57,7 @@ DECLARE_EVENT_CLASS(sched_wakeup_template,
 
 	TP_PROTO(struct task_struct *p, int success),
 
-	TP_ARGS(p, success),
+	TP_ARGS(TP_perf_arg(__task, p), success),
 
 	TP_STRUCT__entry(
 		__array(	char,	comm,	TASK_COMM_LEN	)
@@ -73,9 +73,6 @@ DECLARE_EVENT_CLASS(sched_wakeup_template,
 		__entry->prio		= p->prio;
 		__entry->success	= success;
 		__entry->target_cpu	= task_cpu(p);
-	)
-	TP_perf_assign(
-		__perf_task(p);
 	),
 
 	TP_printk("comm=%s pid=%d prio=%d success=%d target_cpu=%03d",
@@ -313,7 +310,7 @@ DECLARE_EVENT_CLASS(sched_stat_template,
 
 	TP_PROTO(struct task_struct *tsk, u64 delay),
 
-	TP_ARGS(tsk, delay),
+	TP_ARGS(TP_perf_arg(__task, tsk), TP_perf_arg(__count, delay)),
 
 	TP_STRUCT__entry(
 		__array( char,	comm,	TASK_COMM_LEN	)
@@ -325,10 +322,6 @@ DECLARE_EVENT_CLASS(sched_stat_template,
 		memcpy(__entry->comm, tsk->comm, TASK_COMM_LEN);
 		__entry->pid	= tsk->pid;
 		__entry->delay	= delay;
-	)
-	TP_perf_assign(
-		__perf_count(delay);
-		__perf_task(tsk);
 	),
 
 	TP_printk("comm=%s pid=%d delay=%Lu [ns]",
--- a/include/trace/ftrace.h
+++ b/include/trace/ftrace.h
@@ -506,6 +506,9 @@ static inline notrace int ftrace_get_offsets_##call(			\
 #undef TP_perf_assign
 #define TP_perf_assign(args...)
 
+#undef TP_perf_arg
+#define TP_perf_arg(var, val)	(val)
+
 #undef DECLARE_EVENT_CLASS
 #define DECLARE_EVENT_CLASS(call, proto, args, tstruct, assign, print)	\
 									\
@@ -643,6 +646,9 @@ __attribute__((section("_ftrace_events"))) *__event_##call = &event_##call
 #undef TP_perf_assign
 #define TP_perf_assign(args...) args
 
+#undef TP_perf_arg
+#define TP_perf_arg(var, val)	(var = (val))
+
 #undef DECLARE_EVENT_CLASS
 #define DECLARE_EVENT_CLASS(call, proto, args, tstruct, assign, print)	\
 static notrace void							\
@@ -659,13 +665,12 @@ perf_trace_##call(void *__data, proto)					\
 	int __data_size;						\
 	int rctx;							\
 									\
-	perf_fetch_caller_regs(&__regs);				\
-									\
 	__data_size = ftrace_get_offsets_##call(&__data_offsets, args); \
 	__entry_size = ALIGN(__data_size + sizeof(*entry) + sizeof(u32),\
 			     sizeof(u64));				\
 	__entry_size -= sizeof(u32);					\
 									\
+	perf_fetch_caller_regs(&__regs);				\
 	entry = (struct ftrace_raw_##call *)perf_trace_buf_prepare(	\
 		__entry_size, event_call->event.type, &__regs, &rctx);	\
 	if (!entry)							\


  parent reply	other threads:[~2013-06-17 20:22 UTC|newest]

Thread overview: 15+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2013-06-17 17:01 [PATCH 0/3] tracing: more list_empty(perf_events) checks Oleg Nesterov
2013-06-17 17:02 ` [PATCH 1/3] tracing/function: Avoid perf_trace_buf_*() if event_function.perf_events is empty Oleg Nesterov
2013-06-17 17:02 ` [PATCH 2/3] tracing/syscall: Avoid perf_trace_buf_*() if sys_data->perf_events " Oleg Nesterov
2013-06-17 17:02 ` [PATCH 3/3] tracing/perf: Move the PERF_MAX_TRACE_SIZE check into perf_trace_buf_prepare() Oleg Nesterov
2013-06-17 20:18 ` Oleg Nesterov [this message]
2013-06-17 21:27   ` [PATCH 0/3] tracing: more list_empty(perf_events) checks Steven Rostedt
2013-06-18 14:46     ` Oleg Nesterov
2013-06-18 15:41       ` Steven Rostedt
2013-06-18 16:24         ` Oleg Nesterov
2013-07-18  3:00 ` Steven Rostedt
2013-07-18  9:42   ` Peter Zijlstra
2013-07-18 14:39     ` Steven Rostedt
2013-07-18 15:44       ` Oleg Nesterov
2013-07-18 15:53         ` Steven Rostedt
2013-07-18 16:11           ` Oleg Nesterov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20130617201818.GA12349@redhat.com \
    --to=oleg@redhat.com \
    --cc=fweisbec@gmail.com \
    --cc=jovi.zhangwei@huawei.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=masami.hiramatsu.pt@hitachi.com \
    --cc=mingo@redhat.com \
    --cc=rostedt@goodmis.org \
    --cc=srikar@linux.vnet.ibm.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox