ftrace: tracer for scheduler wakeup latency

This patch adds the tracer that tracks the wakeup latency of the
highest priority waking task.

  "wakeup" is added to /debugfs/tracing/available_tracers

Also added to /debugfs/tracing

  tracing_max_latency
     holds the current max latency for the wakeup

  wakeup_thresh
     if set to other than zero, a log will be recorded
     for every wakeup that takes longer than the number
     entered in here (usecs for all counters)
     (deletes previous trace)

Examples:

  (with ftrace_enabled = 0)

============
preemption latency trace v1.1.5 on 2.6.24-rc8
Signed-off-by: Ingo Molnar <mingo@elte.hu>
--------------------------------------------------------------------
 latency: 26 us, #2/2, CPU#1 | (M:rt VP:0, KP:0, SP:0 HP:0 #P:2)
    -----------------
    | task: migration/0-3 (uid:0 nice:-5 policy:1 rt_prio:99)
    -----------------

                 _------=> CPU#
                / _-----=> irqs-off
               | / _----=> need-resched
               || / _---=> hardirq/softirq
               ||| / _--=> preempt-depth
               |||| /
               |||||     delay
   cmd     pid ||||| time  |   caller
      \   /    |||||   \   |   /
   quilt-8551  0d..3    0us+: wake_up_process+0x15/0x17 <ffffffff80233e80> (sched_exec+0xc9/0x100 <ffffffff80235343>)
   quilt-8551  0d..4   26us : sched_switch_callback+0x73/0x81 <ffffffff80338d2f> (schedule+0x483/0x6d5 <ffffffff8048b3ee>)

vim:ft=help
============

  (with ftrace_enabled = 1)

============
preemption latency trace v1.1.5 on 2.6.24-rc8
--------------------------------------------------------------------
 latency: 36 us, #45/45, CPU#0 | (M:rt VP:0, KP:0, SP:0 HP:0 #P:2)
    -----------------
    | task: migration/1-5 (uid:0 nice:-5 policy:1 rt_prio:99)
    -----------------

                 _------=> CPU#
                / _-----=> irqs-off
               | / _----=> need-resched
               || / _---=> hardirq/softirq
               ||| / _--=> preempt-depth
               |||| /
               |||||     delay
   cmd     pid ||||| time  |   caller
      \   /    |||||   \   |   /
    bash-10653 1d..3    0us : wake_up_process+0x15/0x17 <ffffffff80233e80> (sched_exec+0xc9/0x100 <ffffffff80235343>)
    bash-10653 1d..3    1us : try_to_wake_up+0x271/0x2e7 <ffffffff80233dcf> (sub_preempt_count+0xc/0x7a <ffffffff8023309e>)
    bash-10653 1d..2    2us : try_to_wake_up+0x296/0x2e7 <ffffffff80233df4> (update_rq_clock+0x9/0x20 <ffffffff802303f3>)
    bash-10653 1d..2    2us : update_rq_clock+0x1e/0x20 <ffffffff80230408> (__update_rq_clock+0xc/0x90 <ffffffff80230366>)
    bash-10653 1d..2    3us : __update_rq_clock+0x1b/0x90 <ffffffff80230375> (sched_clock+0x9/0x29 <ffffffff80214529>)
    bash-10653 1d..2    4us : try_to_wake_up+0x2a6/0x2e7 <ffffffff80233e04> (activate_task+0xc/0x3f <ffffffff8022ffca>)
    bash-10653 1d..2    4us : activate_task+0x2d/0x3f <ffffffff8022ffeb> (enqueue_task+0xe/0x66 <ffffffff8022ff66>)
    bash-10653 1d..2    5us : enqueue_task+0x5b/0x66 <ffffffff8022ffb3> (enqueue_task_rt+0x9/0x3c <ffffffff80233351>)
    bash-10653 1d..2    6us : try_to_wake_up+0x2ba/0x2e7 <ffffffff80233e18> (check_preempt_wakeup+0x12/0x99 <ffffffff80234f84>)
[...]
    bash-10653 1d..5   33us : tracing_record_cmdline+0xcf/0xd4 <ffffffff80338aad> (_spin_unlock+0x9/0x33 <ffffffff8048d3ec>)
    bash-10653 1d..5   34us : _spin_unlock+0x19/0x33 <ffffffff8048d3fc> (sub_preempt_count+0xc/0x7a <ffffffff8023309e>)
    bash-10653 1d..4   35us : wakeup_sched_switch+0x65/0x2ff <ffffffff80339f66> (_spin_lock_irqsave+0xc/0xa9 <ffffffff8048d08b>)
    bash-10653 1d..4   35us : _spin_lock_irqsave+0x19/0xa9 <ffffffff8048d098> (add_preempt_count+0xe/0x77 <ffffffff8023311a>)
    bash-10653 1d..4   36us : sched_switch_callback+0x73/0x81 <ffffffff80338d2f> (schedule+0x483/0x6d5 <ffffffff8048b3ee>)

vim:ft=help
============

The [...] was added here to not waste your email box space.

Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
diff --git a/kernel/trace/trace_sched_wakeup.c b/kernel/trace/trace_sched_wakeup.c
new file mode 100644
index 0000000..7c3ccef
--- /dev/null
+++ b/kernel/trace/trace_sched_wakeup.c
@@ -0,0 +1,310 @@
+/*
+ * trace task wakeup timings
+ *
+ * Copyright (C) 2007-2008 Steven Rostedt <srostedt@redhat.com>
+ * Copyright (C) 2008 Ingo Molnar <mingo@redhat.com>
+ *
+ * Based on code from the latency_tracer, that is:
+ *
+ *  Copyright (C) 2004-2006 Ingo Molnar
+ *  Copyright (C) 2004 William Lee Irwin III
+ */
+#include <linux/module.h>
+#include <linux/fs.h>
+#include <linux/debugfs.h>
+#include <linux/kallsyms.h>
+#include <linux/uaccess.h>
+#include <linux/ftrace.h>
+
+#include "trace.h"
+
+static struct trace_array	*wakeup_trace;
+static int __read_mostly	tracer_enabled;
+
+static struct task_struct	*wakeup_task;
+static int			wakeup_cpu;
+static unsigned			wakeup_prio = -1;
+
+static DEFINE_SPINLOCK(wakeup_lock);
+
+static void notrace __wakeup_reset(struct trace_array *tr);
+
+/*
+ * Should this new latency be reported/recorded?
+ */
+static int notrace report_latency(cycle_t delta)
+{
+	if (tracing_thresh) {
+		if (delta < tracing_thresh)
+			return 0;
+	} else {
+		if (delta <= tracing_max_latency)
+			return 0;
+	}
+	return 1;
+}
+
+void notrace
+wakeup_sched_switch(struct task_struct *prev, struct task_struct *next)
+{
+	unsigned long latency = 0, t0 = 0, t1 = 0;
+	struct trace_array *tr = wakeup_trace;
+	struct trace_array_cpu *data;
+	cycle_t T0, T1, delta;
+	unsigned long flags;
+	long disabled;
+	int cpu;
+
+	if (unlikely(!tracer_enabled))
+		return;
+
+	/*
+	 * When we start a new trace, we set wakeup_task to NULL
+	 * and then set tracer_enabled = 1. We want to make sure
+	 * that another CPU does not see the tracer_enabled = 1
+	 * and the wakeup_task with an older task, that might
+	 * actually be the same as next.
+	 */
+	smp_rmb();
+
+	if (next != wakeup_task)
+		return;
+
+	/* The task we are waitng for is waking up */
+	data = tr->data[wakeup_cpu];
+
+	/* disable local data, not wakeup_cpu data */
+	cpu = raw_smp_processor_id();
+	disabled = atomic_inc_return(&tr->data[cpu]->disabled);
+	if (likely(disabled != 1))
+		goto out;
+
+	spin_lock_irqsave(&wakeup_lock, flags);
+
+	/* We could race with grabbing wakeup_lock */
+	if (unlikely(!tracer_enabled || next != wakeup_task))
+		goto out_unlock;
+
+	ftrace(tr, data, CALLER_ADDR1, CALLER_ADDR2, flags);
+
+	/*
+	 * usecs conversion is slow so we try to delay the conversion
+	 * as long as possible:
+	 */
+	T0 = data->preempt_timestamp;
+	T1 = now(cpu);
+	delta = T1-T0;
+
+	if (!report_latency(delta))
+		goto out_unlock;
+
+	latency = nsecs_to_usecs(delta);
+
+	tracing_max_latency = delta;
+	t0 = nsecs_to_usecs(T0);
+	t1 = nsecs_to_usecs(T1);
+
+	update_max_tr(tr, wakeup_task, wakeup_cpu);
+
+	if (tracing_thresh) {
+		printk(KERN_INFO "(%16s-%-5d|#%d): %lu us wakeup latency "
+		       "violates %lu us threshold.\n"
+		       " => started at timestamp %lu: ",
+				wakeup_task->comm, wakeup_task->pid,
+				raw_smp_processor_id(),
+				latency, nsecs_to_usecs(tracing_thresh), t0);
+	} else {
+		printk(KERN_INFO "(%16s-%-5d|#%d): new %lu us maximum "
+		       "wakeup latency.\n => started at timestamp %lu: ",
+				wakeup_task->comm, wakeup_task->pid,
+				cpu, latency, t0);
+	}
+
+	printk(KERN_CONT "   ended at timestamp %lu: ", t1);
+	dump_stack();
+	t1 = nsecs_to_usecs(now(cpu));
+	printk(KERN_CONT "   dump-end timestamp %lu\n\n", t1);
+
+out_unlock:
+	__wakeup_reset(tr);
+	spin_unlock_irqrestore(&wakeup_lock, flags);
+out:
+	atomic_dec(&tr->data[cpu]->disabled);
+}
+
+static void notrace __wakeup_reset(struct trace_array *tr)
+{
+	struct trace_array_cpu *data;
+	int cpu;
+
+	assert_spin_locked(&wakeup_lock);
+
+	for_each_possible_cpu(cpu) {
+		data = tr->data[cpu];
+		tracing_reset(data);
+	}
+
+	wakeup_cpu = -1;
+	wakeup_prio = -1;
+
+	if (wakeup_task)
+		put_task_struct(wakeup_task);
+
+	wakeup_task = NULL;
+}
+
+static void notrace wakeup_reset(struct trace_array *tr)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&wakeup_lock, flags);
+	__wakeup_reset(tr);
+	spin_unlock_irqrestore(&wakeup_lock, flags);
+}
+
+static notrace void
+wakeup_check_start(struct trace_array *tr, struct task_struct *p,
+		   struct task_struct *curr)
+{
+	int cpu = smp_processor_id();
+	unsigned long flags;
+	long disabled;
+
+	if (likely(!rt_task(p)) ||
+			p->prio >= wakeup_prio ||
+			p->prio >= curr->prio)
+		return;
+
+	disabled = atomic_inc_return(&tr->data[cpu]->disabled);
+	if (unlikely(disabled != 1))
+		goto out;
+
+	/* interrupts should be off from try_to_wake_up */
+	spin_lock(&wakeup_lock);
+
+	/* check for races. */
+	if (!tracer_enabled || p->prio >= wakeup_prio)
+		goto out_locked;
+
+	/* reset the trace */
+	__wakeup_reset(tr);
+
+	wakeup_cpu = task_cpu(p);
+	wakeup_prio = p->prio;
+
+	wakeup_task = p;
+	get_task_struct(wakeup_task);
+
+	local_save_flags(flags);
+
+	tr->data[wakeup_cpu]->preempt_timestamp = now(cpu);
+	ftrace(tr, tr->data[wakeup_cpu], CALLER_ADDR1, CALLER_ADDR2, flags);
+
+out_locked:
+	spin_unlock(&wakeup_lock);
+out:
+	atomic_dec(&tr->data[cpu]->disabled);
+}
+
+notrace void
+ftrace_wake_up_task(struct task_struct *wakee, struct task_struct *curr)
+{
+	if (likely(!tracer_enabled))
+		return;
+
+	wakeup_check_start(wakeup_trace, wakee, curr);
+}
+
+notrace void
+ftrace_wake_up_new_task(struct task_struct *wakee, struct task_struct *curr)
+{
+	if (likely(!tracer_enabled))
+		return;
+
+	wakeup_check_start(wakeup_trace, wakee, curr);
+}
+
+static notrace void start_wakeup_tracer(struct trace_array *tr)
+{
+	wakeup_reset(tr);
+
+	/*
+	 * Don't let the tracer_enabled = 1 show up before
+	 * the wakeup_task is reset. This may be overkill since
+	 * wakeup_reset does a spin_unlock after setting the
+	 * wakeup_task to NULL, but I want to be safe.
+	 * This is a slow path anyway.
+	 */
+	smp_wmb();
+
+	tracer_enabled = 1;
+
+	return;
+}
+
+static notrace void stop_wakeup_tracer(struct trace_array *tr)
+{
+	tracer_enabled = 0;
+}
+
+static notrace void wakeup_tracer_init(struct trace_array *tr)
+{
+	wakeup_trace = tr;
+
+	if (tr->ctrl)
+		start_wakeup_tracer(tr);
+}
+
+static notrace void wakeup_tracer_reset(struct trace_array *tr)
+{
+	if (tr->ctrl) {
+		stop_wakeup_tracer(tr);
+		/* make sure we put back any tasks we are tracing */
+		wakeup_reset(tr);
+	}
+}
+
+static void wakeup_tracer_ctrl_update(struct trace_array *tr)
+{
+	if (tr->ctrl)
+		start_wakeup_tracer(tr);
+	else
+		stop_wakeup_tracer(tr);
+}
+
+static void notrace wakeup_tracer_open(struct trace_iterator *iter)
+{
+	/* stop the trace while dumping */
+	if (iter->tr->ctrl)
+		stop_wakeup_tracer(iter->tr);
+}
+
+static void notrace wakeup_tracer_close(struct trace_iterator *iter)
+{
+	/* forget about any processes we were recording */
+	if (iter->tr->ctrl)
+		start_wakeup_tracer(iter->tr);
+}
+
+static struct tracer wakeup_tracer __read_mostly =
+{
+	.name		= "wakeup",
+	.init		= wakeup_tracer_init,
+	.reset		= wakeup_tracer_reset,
+	.open		= wakeup_tracer_open,
+	.close		= wakeup_tracer_close,
+	.ctrl_update	= wakeup_tracer_ctrl_update,
+	.print_max	= 1,
+};
+
+__init static int init_wakeup_tracer(void)
+{
+	int ret;
+
+	ret = register_tracer(&wakeup_tracer);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+device_initcall(init_wakeup_tracer);