ftrace: add wakeup events to sched tracer

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

5 files changed, 88 insertions, 5 deletions

diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
index f589805..192c135 100644
--- a/kernel/trace/trace.c
+++ b/kernel/trace/trace.c
@@ -57,7 +57,7 @@ static struct trace_array	max_tr;
 static DEFINE_PER_CPU(struct trace_array_cpu, max_data);
 
 static int			tracer_enabled = 1;
-static unsigned long		trace_nr_entries = 16384UL;
+static unsigned long		trace_nr_entries = 65536UL;
 
 static struct tracer		*trace_types __read_mostly;
 static struct tracer		*current_trace __read_mostly;
@@ -87,6 +87,7 @@ enum trace_type {
 
 	TRACE_FN,
 	TRACE_CTX,
+	TRACE_WAKE,
 	TRACE_SPECIAL,
 
 	__TRACE_LAST_TYPE
@@ -711,6 +712,30 @@ tracing_sched_switch_trace(struct trace_array *tr,
 		wake_up (&trace_wait);
 }
 
+void
+tracing_sched_wakeup_trace(struct trace_array *tr,
+			   struct trace_array_cpu *data,
+			   struct task_struct *wakee, struct task_struct *curr,
+			   unsigned long flags)
+{
+	struct trace_entry *entry;
+	unsigned long irq_flags;
+
+	spin_lock_irqsave(&data->lock, irq_flags);
+	entry			= tracing_get_trace_entry(tr, data);
+	tracing_generic_entry_update(entry, flags);
+	entry->type		= TRACE_WAKE;
+	entry->ctx.prev_pid	= curr->pid;
+	entry->ctx.prev_prio	= curr->prio;
+	entry->ctx.prev_state	= curr->state;
+	entry->ctx.next_pid	= wakee->pid;
+	entry->ctx.next_prio	= wakee->prio;
+	spin_unlock_irqrestore(&data->lock, irq_flags);
+
+	if (!(trace_flags & TRACE_ITER_BLOCK))
+		wake_up(&trace_wait);
+}
+
 #ifdef CONFIG_FTRACE
 static void
 function_trace_call(unsigned long ip, unsigned long parent_ip)
@@ -1183,13 +1208,14 @@ print_lat_fmt(struct trace_iterator *iter, unsigned int trace_idx, int cpu)
 		trace_seq_puts(s, ")\n");
 		break;
 	case TRACE_CTX:
+	case TRACE_WAKE:
 		S = entry->ctx.prev_state < sizeof(state_to_char) ?
 			state_to_char[entry->ctx.prev_state] : 'X';
 		comm = trace_find_cmdline(entry->ctx.next_pid);
-		trace_seq_printf(s, " %d:%d:%c --> %d:%d %s\n",
+		trace_seq_printf(s, " %5d:%3d:%c %s %5d:%3d %s\n",
 				 entry->ctx.prev_pid,
 				 entry->ctx.prev_prio,
-				 S,
+				 S, entry->type == TRACE_CTX ? "==>" : "  +",
 				 entry->ctx.next_pid,
 				 entry->ctx.next_prio,
 				 comm);
@@ -1256,12 +1282,14 @@ static int print_trace_fmt(struct trace_iterator *iter)
 			return 0;
 		break;
 	case TRACE_CTX:
+	case TRACE_WAKE:
 		S = entry->ctx.prev_state < sizeof(state_to_char) ?
 			state_to_char[entry->ctx.prev_state] : 'X';
-		ret = trace_seq_printf(s, " %d:%d:%c ==> %d:%d\n",
+		ret = trace_seq_printf(s, " %5d:%3d:%c %s %5d:%3d\n",
 				       entry->ctx.prev_pid,
 				       entry->ctx.prev_prio,
 				       S,
+				       entry->type == TRACE_CTX ? "==>" : "  +",
 				       entry->ctx.next_pid,
 				       entry->ctx.next_prio);
 		if (!ret)
@@ -1301,8 +1329,11 @@ static int print_raw_fmt(struct trace_iterator *iter)
 			return 0;
 		break;
 	case TRACE_CTX:
+	case TRACE_WAKE:
 		S = entry->ctx.prev_state < sizeof(state_to_char) ?
 			state_to_char[entry->ctx.prev_state] : 'X';
+		if (entry->type == TRACE_WAKE)
+			S = '+';
 		ret = trace_seq_printf(s, "%d %d %c %d %d\n",
 				       entry->ctx.prev_pid,
 				       entry->ctx.prev_prio,
@@ -1355,8 +1386,11 @@ static int print_hex_fmt(struct trace_iterator *iter)
 		SEQ_PUT_HEX_FIELD_RET(s, entry->fn.parent_ip);
 		break;
 	case TRACE_CTX:
+	case TRACE_WAKE:
 		S = entry->ctx.prev_state < sizeof(state_to_char) ?
 			state_to_char[entry->ctx.prev_state] : 'X';
+		if (entry->type == TRACE_WAKE)
+			S = '+';
 		SEQ_PUT_HEX_FIELD_RET(s, entry->ctx.prev_pid);
 		SEQ_PUT_HEX_FIELD_RET(s, entry->ctx.prev_prio);
 		SEQ_PUT_HEX_FIELD_RET(s, S);
diff --git a/kernel/trace/trace.h b/kernel/trace/trace.h
index 2b7352b..90e0ba0 100644
--- a/kernel/trace/trace.h
+++ b/kernel/trace/trace.h
@@ -164,6 +164,12 @@ void tracing_sched_switch_trace(struct trace_array *tr,
 				struct task_struct *next,
 				unsigned long flags);
 void tracing_record_cmdline(struct task_struct *tsk);
+
+void tracing_sched_wakeup_trace(struct trace_array *tr,
+				struct trace_array_cpu *data,
+				struct task_struct *wakee,
+				struct task_struct *cur,
+				unsigned long flags);
 void trace_special(struct trace_array *tr,
 		   struct trace_array_cpu *data,
 		   unsigned long arg1,
@@ -194,11 +200,17 @@ extern cycle_t ftrace_now(int cpu);
 #ifdef CONFIG_SCHED_TRACER
 extern void
 wakeup_sched_switch(struct task_struct *prev, struct task_struct *next);
+extern void
+wakeup_sched_wakeup(struct task_struct *wakee, struct task_struct *curr);
 #else
 static inline void
 wakeup_sched_switch(struct task_struct *prev, struct task_struct *next)
 {
 }
+static inline void
+wakeup_sched_wakeup(struct task_struct *wakee, struct task_struct *curr)
+{
+}
 #endif
 
 #ifdef CONFIG_CONTEXT_SWITCH_TRACER
diff --git a/kernel/trace/trace_sched_switch.c b/kernel/trace/trace_sched_switch.c
index b738eac..8b1cf1a 100644
--- a/kernel/trace/trace_sched_switch.c
+++ b/kernel/trace/trace_sched_switch.c
@@ -41,6 +41,29 @@ ctx_switch_func(struct task_struct *prev, struct task_struct *next)
 	local_irq_restore(flags);
 }
 
+static void wakeup_func(struct task_struct *wakee, struct task_struct *curr)
+{
+	struct trace_array *tr = ctx_trace;
+	struct trace_array_cpu *data;
+	unsigned long flags;
+	long disabled;
+	int cpu;
+
+	if (!tracer_enabled)
+		return;
+
+	local_irq_save(flags);
+	cpu = raw_smp_processor_id();
+	data = tr->data[cpu];
+	disabled = atomic_inc_return(&data->disabled);
+
+	if (likely(disabled == 1))
+		tracing_sched_wakeup_trace(tr, data, wakee, curr, flags);
+
+	atomic_dec(&data->disabled);
+	local_irq_restore(flags);
+}
+
 void ftrace_ctx_switch(struct task_struct *prev, struct task_struct *next)
 {
 	tracing_record_cmdline(prev);
@@ -57,6 +80,19 @@ void ftrace_ctx_switch(struct task_struct *prev, struct task_struct *next)
 	wakeup_sched_switch(prev, next);
 }
 
+void
+ftrace_wake_up_task(struct task_struct *wakee, struct task_struct *curr)
+{
+	tracing_record_cmdline(curr);
+
+	wakeup_func(wakee, curr);
+
+	/*
+	 * Chain to the wakeup tracer (this is a NOP if disabled):
+	 */
+	wakeup_sched_wakeup(wakee, curr);
+}
+
 static void sched_switch_reset(struct trace_array *tr)
 {
 	int cpu;
diff --git a/kernel/trace/trace_sched_wakeup.c b/kernel/trace/trace_sched_wakeup.c
index 662679c..87fa7b2 100644
--- a/kernel/trace/trace_sched_wakeup.c
+++ b/kernel/trace/trace_sched_wakeup.c
@@ -202,7 +202,7 @@ out:
 }
 
 void
-ftrace_wake_up_task(struct task_struct *wakee, struct task_struct *curr)
+wakeup_sched_wakeup(struct task_struct *wakee, struct task_struct *curr)
 {
 	if (likely(!tracer_enabled))
 		return;
diff --git a/kernel/trace/trace_selftest.c b/kernel/trace/trace_selftest.c
index 85715b8..39dd452 100644
--- a/kernel/trace/trace_selftest.c
+++ b/kernel/trace/trace_selftest.c
@@ -8,6 +8,7 @@ static inline int trace_valid_entry(struct trace_entry *entry)
 	switch (entry->type) {
 	case TRACE_FN:
 	case TRACE_CTX:
+	case TRACE_WAKE:
 		return 1;
 	}
 	return 0;