ftrace perf: Fixes and speedup

Currently we check sample type for ftrace:function event even if it's not created as sampling event. That prevents creating ftrace_function event in counting mode.

Making sure we check sample types only for sampling events.

Before:

  $ sudo perf stat -e ftrace:function ls
  ...

Performance counter stats for 'ls':

<not supported> ftrace:function

0.001983662 seconds time elapsed

After:

  $ sudo perf stat -e ftrace:function ls
  ...

   Performance counter stats for 'ls':

              44,498      ftrace:function

         0.037534722 seconds time elapsed

Signed-off-by Jiri Olsa <[email protected]>

---
 kernel/trace/trace_event_perf.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

kernel/trace/trace_event_perf.c +1 -1

index 00df25fd86ef..a7171ec2c1ca 100644
@@ -52,14 +52,14 @@ static int perf_trace_event_perm(struct trace_event_call *tp_event,
 		 * event, due to issues with page faults while tracing page
 		 * fault handler and its overall trickiness nature.
 		 */
-		if (!p_event->attr.exclude_callchain_user)
+		if (is_sampling_event(p_event) && !p_event->attr.exclude_callchain_user)
 			return -EINVAL;

Hi Jiri,

Quoted 43 lines · Jiri Olsa

Currently we check sample type for ftrace:function event even if it's not created as sampling event. That prevents creating ftrace_function event in counting mode.

Making sure we check sample types only for sampling events.

Before:

  $ sudo perf stat -e ftrace:function ls
  ...

Performance counter stats for 'ls':

<not supported> ftrace:function

0.001983662 seconds time elapsed

After:

  $ sudo perf stat -e ftrace:function ls
  ...

   Performance counter stats for 'ls':

              44,498      ftrace:function

         0.037534722 seconds time elapsed

Signed-off-by Jiri Olsa <[email protected]>

---
 kernel/trace/trace_event_perf.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

kernel/trace/trace_event_perf.c +1 -1

index 00df25fd86ef..a7171ec2c1ca 100644
@@ -52,14 +52,14 @@ static int perf_trace_event_perm(struct trace_event_call *tp_event,
 		 * event, due to issues with page faults while tracing page
 		 * fault handler and its overall trickiness nature.
 		 */
-		if (!p_event->attr.exclude_callchain_user)
+		if (is_sampling_event(p_event) && !p_event->attr.exclude_callchain_user)
 			return -EINVAL;
 

kernel/trace/trace_event_perf.c static int perf_trace_event_perm(struct trace_event_call *tp_event, · line 65

Quoted 9 lines

kernel/trace/trace_event_perf.c +1 -1

 		/*
 		 * Same reason to disable user stack dump as for user space
 		 * callchains above.
 		 */
-		if (p_event->attr.sample_type & PERF_SAMPLE_STACK_USER)
+		if (is_sampling_event(p_event) && p_event->attr.sample_type & PERF_SAMPLE_STACK_USER)
 			return -EINVAL;
 	}
 

What about checking is_sampling_event() first and goto the last paranoid_tracepoint_raw check instead? This way we can remove the same check in the function trace case.

Quoted 46 lines · Namhyung Kim

Hi Jiri,

Quoted 43 lines · Jiri Olsa

Currently we check sample type for ftrace:function event even if it's not created as sampling event. That prevents creating ftrace_function event in counting mode.

Making sure we check sample types only for sampling events.

Before:

  $ sudo perf stat -e ftrace:function ls
  ...

Performance counter stats for 'ls':

<not supported> ftrace:function

0.001983662 seconds time elapsed

After:

  $ sudo perf stat -e ftrace:function ls
  ...

   Performance counter stats for 'ls':

              44,498      ftrace:function

         0.037534722 seconds time elapsed

Signed-off-by Jiri Olsa <[email protected]>

---
 kernel/trace/trace_event_perf.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

kernel/trace/trace_event_perf.c +1 -1

index 00df25fd86ef..a7171ec2c1ca 100644
@@ -52,14 +52,14 @@ static int perf_trace_event_perm(struct trace_event_call *tp_event,
 		 * event, due to issues with page faults while tracing page
 		 * fault handler and its overall trickiness nature.
 		 */
-		if (!p_event->attr.exclude_callchain_user)
+		if (is_sampling_event(p_event) && !p_event->attr.exclude_callchain_user)
 			return -EINVAL;
 

Quoted 13 lines

kernel/trace/trace_event_perf.c static int perf_trace_event_perm(struct trace_event_call *tp_event, · line 65

Quoted 9 lines

kernel/trace/trace_event_perf.c +1 -1

 		/*
 		 * Same reason to disable user stack dump as for user space
 		 * callchains above.
 		 */
-		if (p_event->attr.sample_type & PERF_SAMPLE_STACK_USER)
+		if (is_sampling_event(p_event) && p_event->attr.sample_type & PERF_SAMPLE_STACK_USER)
 			return -EINVAL;
 	}
 

What about checking is_sampling_event() first and goto the last paranoid_tracepoint_raw check instead? This way we can remove the same check in the function trace case.

right, will check

Quoted 61 lines · Jiri Olsa

Quoted 59 lines · Namhyung Kim

Hi Jiri,

Quoted 42 lines · Jiri Olsa

Currently we check sample type for ftrace:function event even if it's not created as sampling event. That prevents creating ftrace_function event in counting mode.

Making sure we check sample types only for sampling events.

Before:

  $ sudo perf stat -e ftrace:function ls
  ...

Performance counter stats for 'ls':

<not supported> ftrace:function

0.001983662 seconds time elapsed

After:

  $ sudo perf stat -e ftrace:function ls
  ...

   Performance counter stats for 'ls':

              44,498      ftrace:function

         0.037534722 seconds time elapsed

Signed-off-by Jiri Olsa <[email protected]>

---
 kernel/trace/trace_event_perf.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

kernel/trace/trace_event_perf.c +1 -1

index 00df25fd86ef..a7171ec2c1ca 100644
@@ -52,14 +52,14 @@ static int perf_trace_event_perm(struct trace_event_call *tp_event,
 		 * event, due to issues with page faults while tracing page
 		 * fault handler and its overall trickiness nature.
 		 */
-		if (!p_event->attr.exclude_callchain_user)
+		if (is_sampling_event(p_event) && !p_event->attr.exclude_callchain_user)
 			return -EINVAL;

kernel/trace/trace_event_perf.c static int perf_trace_event_perm(struct trace_event_call *tp_event, · line 65

Quoted 10 lines

kernel/trace/trace_event_perf.c +1 -1

 
 		/*
 		 * Same reason to disable user stack dump as for user space
 		 * callchains above.
 		 */
-		if (p_event->attr.sample_type & PERF_SAMPLE_STACK_USER)
+		if (is_sampling_event(p_event) && p_event->attr.sample_type & PERF_SAMPLE_STACK_USER)
 			return -EINVAL;
 	}
 

What about checking is_sampling_event() first and goto the last paranoid_tracepoint_raw check instead? This way we can remove the same check in the function trace case.

Quoted 1 line

right, will check

hum, did you mean something like this?

I'd rather keep it the original way.. seems more straight

Quoted 4 lines

Quoted 4 lines

Quoted 3 lines

What about checking is_sampling_event() first and goto the last paranoid_tracepoint_raw check instead? This way we can remove the same check in the function trace case.

Quoted 5 lines

Quoted 1 line

right, will check

hum, did you mean something like this?

I'd rather keep it the original way.. seems more straight

Hmm.. I think I was wrong. But it seems we can simply return 0 for non sampling case. How about this?

Currently we check sample type for ftrace:function event even if it's not created as sampling event. That prevents creating ftrace_function event in counting mode.

Making sure we check sample types only for sampling events.

Before:

  $ sudo perf stat -e ftrace:function ls
  ...

Performance counter stats for 'ls':

<not supported> ftrace:function

0.001983662 seconds time elapsed

After:

  $ sudo perf stat -e ftrace:function ls

Quoted 14 lines · On Wed, 9 Mar 2016 21:46:41 +0100 Jiri Olsa <[email protected]> wrote:

Currently we check sample type for ftrace:function event even if it's not created as sampling event. That prevents creating ftrace_function event in counting mode.

Making sure we check sample types only for sampling events.

Before:

  $ sudo perf stat -e ftrace:function ls
  ...

Performance counter stats for 'ls':

<not supported> ftrace:function

Quoted 4 lines

         0.001983662 seconds time elapsed

After:

  $ sudo perf stat -e ftrace:function ls

I'm assuming you gave yourself admin capabilities, and not any normal user may sample function tracing, right?

We suppress events with attr::exclude_kernel set when the event is generated, so following capture will give no warning but won't produce any data:

  $ sudo perf record -e sched:sched_switch:u ls
  $ sudo /perf script | wc -l
  0

Checking the attr::exclude_(kernel|user) at the event init time and failing right away for tracepoints from uprobes/kprobes and native ones:

  $ sudo perf record -e sched:sched_switch:u ls
  Error:
  The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (sched:sched_switch).
  /bin/dmesg may provide additional information.
  No CONFIG_PERF_EVENTS=y kernel support configured?

  $ sudo perf record -e probe:sys_read:u ls
  Error:
  The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (probe:sys_read).
  /bin/dmesg may provide additional information.
  No CONFIG_PERF_EVENTS=y kernel support configured?

  $ ./perf record -e probe_ex:main:k ./ex
  Error:
  The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (probe_ex:main).
  /bin/dmesg may provide additional information.
  No CONFIG_PERF_EVENTS=y kernel support configured?

Signed-off-by Jiri Olsa <[email protected]>

---
 kernel/events/core.c            |  5 -----
 kernel/trace/trace_event_perf.c | 25 +++++++++++++++++++++++++
 2 files changed, 25 insertions(+), 5 deletions(-)

kernel/events/core.c +0 -5

index c15fd097af93..ca68fdcf47ce 100644
@@ -6898,11 +6898,6 @@ static int perf_tp_event_match(struct perf_event *event,
 {
 	if (event->hw.state & PERF_HES_STOPPED)
 		return 0;
-	/*
-	 * All tracepoints are from kernel-space.
-	 */
-	if (event->attr.exclude_kernel)
-		return 0;
 
 	if (!perf_tp_filter_match(event, data))
 		return 0;

kernel/trace/trace_event_perf.c +25 -0

index a7171ec2c1ca..0a3779bd18a1 100644
@@ -182,11 +182,36 @@ static void perf_trace_event_close(struct perf_event *p_event)
 	tp_event->class->reg(tp_event, TRACE_REG_PERF_CLOSE, p_event);
 }
 
+static int perf_trace_event_attr(struct trace_event_call *tp_event,
+				 struct perf_event *event)
+{
+	/*
+	 * All tracepoints and kprobes are from kernel-space.
+	 */
+	if (((tp_event->flags & TRACE_EVENT_FL_TRACEPOINT) ||
+	     (tp_event->flags & TRACE_EVENT_FL_KPROBE)) &&
+	     event->attr.exclude_kernel)
+		return -EINVAL;
+
+	/*
+	 * All uprobes are from user-space.
+	 */
+	if ((tp_event->flags & TRACE_EVENT_FL_UPROBE) &&
+	    event->attr.exclude_user)
+		return -EINVAL;
+
+	return 0;
+}
+
 static int perf_trace_event_init(struct trace_event_call *tp_event,
 				 struct perf_event *p_event)
 {
 	int ret;
 
+	ret = perf_trace_event_attr(tp_event, p_event);
+	if (ret)
+		return ret;
+
 	ret = perf_trace_event_perm(tp_event, p_event);
 	if (ret)
 		return ret;

Signature · 1 line

2.4.3

Quoted 30 lines · Jiri Olsa

We suppress events with attr::exclude_kernel set when the event is generated, so following capture will give no warning but won't produce any data:

  $ sudo perf record -e sched:sched_switch:u ls
  $ sudo /perf script | wc -l
  0

Checking the attr::exclude_(kernel|user) at the event init time and failing right away for tracepoints from uprobes/kprobes and native ones:

  $ sudo perf record -e sched:sched_switch:u ls
  Error:
  The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (sched:sched_switch).
  /bin/dmesg may provide additional information.
  No CONFIG_PERF_EVENTS=y kernel support configured?

  $ sudo perf record -e probe:sys_read:u ls
  Error:
  The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (probe:sys_read).
  /bin/dmesg may provide additional information.
  No CONFIG_PERF_EVENTS=y kernel support configured?

  $ ./perf record -e probe_ex:main:k ./ex
  Error:
  The sys_perf_event_open() syscall returned with 22 (Invalid argument) for event (probe_ex:main).
  /bin/dmesg may provide additional information.
  No CONFIG_PERF_EVENTS=y kernel support configured?

Quoted 1 line

Signed-off-by Jiri Olsa <[email protected]>

Acked-by Namhyung Kim <[email protected]>

Maybe we need to improve the error message later.

Having following commands running concurrently:

  # perf record -e ftrace:function -a -o krava.data sleep 10
  # perf record -e ftrace:function --filter 'ip == SyS_read' ls

will end up in the latter one to fail on the filter rules and store all functions (in perf.data) as instructed by the first perf record instead of just SyS_read records.

The reason for this is, that tracepoint code by default triggers all events that registered for the tracepoint.

While ftrace:function is special because ftrace_ops itself carries a filter and only the event that owns ftrace_ops is eligible to be triggered.

Fixing this by using ftrace_ops::private value to keep the perf_event pointer. This way we don't need to search for triggered event (as tracepoint handler does) and directly store sample.

Suggested-by Steven Rostedt <[email protected]>

Signed-off-by Jiri Olsa <[email protected]>

---
 include/linux/perf_event.h      |  3 +++
 kernel/events/core.c            | 22 ++++++++++++++++++++++
 kernel/trace/trace_event_perf.c | 10 +++-------
 3 files changed, 28 insertions(+), 7 deletions(-)

include/linux/perf_event.h +3 -0

index a9d8cab18b00..a330dc06d90d 100644
@@ -1008,6 +1008,9 @@ extern void perf_tp_event(u64 addr, u64 count, void *record,
 			  int entry_size, struct pt_regs *regs,
 			  struct hlist_head *head, int rctx,
 			  struct task_struct *task);
+void perf_function_event(struct perf_event *event,
+			 void *record, int entry_size,
+			 struct pt_regs *regs);
 extern void perf_bp_event(struct perf_event *event, void *data);
 
 #ifndef perf_misc_flags

kernel/events/core.c +22 -0

index ca68fdcf47ce..18da90859c17 100644
@@ -7063,6 +7063,28 @@ static void perf_event_free_bpf_prog(struct perf_event *event)
 	}
 }
 
+#ifdef CONFIG_FUNCTION_TRACER
+void perf_function_event(struct perf_event *event,
+			 void *record, int entry_size,
+			 struct pt_regs *regs)
+
+{
+	struct perf_sample_data data;
+	struct perf_raw_record raw = {
+		.size = entry_size,
+		.data = record,
+	};
+
+	if (event->hw.state & PERF_HES_STOPPED)
+		return;
+
+	perf_sample_data_init(&data, 0, 0);
+	data.raw = &raw;
+
+	perf_swevent_event(event, 1, &data, regs);
+}
+#endif /* CONFIG_FUNCTION_TRACER */
+
 #else
 
 static inline void perf_tp_register(void)

kernel/trace/trace_event_perf.c +3 -7

index 0a3779bd18a1..087c811db04c 100644
@@ -328,14 +328,9 @@ perf_ftrace_function_call(unsigned long ip, unsigned long parent_ip,
 			  struct ftrace_ops *ops, struct pt_regs *pt_regs)
 {
 	struct ftrace_entry *entry;
-	struct hlist_head *head;
 	struct pt_regs regs;
 	int rctx;
 
-	head = this_cpu_ptr(event_function.perf_events);
-	if (hlist_empty(head))
-		return;
-
 #define ENTRY_SIZE (ALIGN(sizeof(struct ftrace_entry) + sizeof(u32), \
 		    sizeof(u64)) - sizeof(u32))
 

@@ -349,8 +344,8 @@ perf_ftrace_function_call(unsigned long ip, unsigned long parent_ip,
 
 	entry->ip = ip;
 	entry->parent_ip = parent_ip;
-	perf_trace_buf_submit(entry, ENTRY_SIZE, rctx, 0,
-			      1, &regs, head, NULL);
+	perf_function_event(ops->private, entry, ENTRY_SIZE, &regs);
+	perf_swevent_put_recursion_context(rctx);
 
 #undef ENTRY_SIZE
 }

@@ -359,6 +354,7 @@ static int perf_ftrace_function_register(struct perf_event *event)
 {
 	struct ftrace_ops *ops = &event->ftrace_ops;
 
+	ops->private = event;
 	ops->flags |= FTRACE_OPS_FL_PER_CPU | FTRACE_OPS_FL_RCU;
 	ops->func = perf_ftrace_function_call;
 	return register_ftrace_function(ops);

Signature · 1 line

2.4.3

Change __ftrace_hash_rec_update to return true in case we need to update dynamic ftrace call records. It return false in case no update is needed.

Signed-off-by Jiri Olsa <[email protected]>

---
 kernel/trace/ftrace.c | 26 ++++++++++++++++----------
 1 file changed, 16 insertions(+), 10 deletions(-)

kernel/trace/ftrace.c +6 -5

index eca592f977b2..123dddc660e9 100644
@@ -1610,7 +1610,7 @@ static bool test_rec_ops_needs_regs(struct dyn_ftrace *rec)
 	return  keep_regs;
 }
 
-static void __ftrace_hash_rec_update(struct ftrace_ops *ops,
+static bool __ftrace_hash_rec_update(struct ftrace_ops *ops,
 				     int filter_hash,
 				     bool inc)
 {

@@ -1618,12 +1618,13 @@ static void __ftrace_hash_rec_update(struct ftrace_ops *ops,
 	struct ftrace_hash *other_hash;
 	struct ftrace_page *pg;
 	struct dyn_ftrace *rec;
+	bool update = false;
 	int count = 0;
 	int all = 0;
 
 	/* Only update if the ops has been registered */
 	if (!(ops->flags & FTRACE_OPS_FL_ENABLED))
-		return;
+		return false;
 
 	/*
 	 * In the filter_hash case:

@@ -1650,7 +1651,7 @@ static void __ftrace_hash_rec_update(struct ftrace_ops *ops,
 		 * then there's nothing to do.
 		 */
 		if (ftrace_hash_empty(hash))
-			return;
+			return false;
 	}
 
 	do_for_each_ftrace_rec(pg, rec) {

@@ -1694,7 +1695,7 @@ static void __ftrace_hash_rec_update(struct ftrace_ops *ops,
 		if (inc) {
 			rec->flags++;
 			if (FTRACE_WARN_ON(ftrace_rec_count(rec) == FTRACE_REF_MAX))
-				return;
+				return false;
 
 			/*
 			 * If there's only a single callback registered to a

@@ -1720,7 +1721,7 @@ static void __ftrace_hash_rec_update(struct ftrace_ops *ops,
 				rec->flags |= FTRACE_FL_REGS;
 		} else {
 			if (FTRACE_WARN_ON(ftrace_rec_count(rec) == 0))
-				return;
+				return false;
 			rec->flags--;
 
 			/*

kernel/trace/ftrace.c +8 -5

+
 		/* Shortcut, if we handled all records, we are done. */
 		if (!all && count == hash->count)
-			return;
+			return update;
 	} while_for_each_ftrace_rec();
+
+	return update;
 }
 
-static void ftrace_hash_rec_disable(struct ftrace_ops *ops,
+static bool ftrace_hash_rec_disable(struct ftrace_ops *ops,
 				    int filter_hash)
 {
-	__ftrace_hash_rec_update(ops, filter_hash, 0);
+	return __ftrace_hash_rec_update(ops, filter_hash, 0);
 }
 
-static void ftrace_hash_rec_enable(struct ftrace_ops *ops,
+static bool ftrace_hash_rec_enable(struct ftrace_ops *ops,
 				   int filter_hash)
 {
-	__ftrace_hash_rec_update(ops, filter_hash, 1);
+	return __ftrace_hash_rec_update(ops, filter_hash, 1);
 }
 
 static void ftrace_hash_rec_update_modify(struct ftrace_ops *ops,

Signature · 1 line

2.4.3

Quoted 17 lines

kernel/trace/ftrace.c +2 -2

@@ -1694,7 +1695,7 @@ static void __ftrace_hash_rec_update(struct ftrace_ops *ops,
 		if (inc) {
 			rec->flags++;
 			if (FTRACE_WARN_ON(ftrace_rec_count(rec) == FTRACE_REF_MAX))
-				return;
+				return false;
 
 			/*
 			 * If there's only a single callback registered to a

@@ -1720,7 +1721,7 @@ static void __ftrace_hash_rec_update(struct ftrace_ops *ops,
 				rec->flags |= FTRACE_FL_REGS;
 		} else {
 			if (FTRACE_WARN_ON(ftrace_rec_count(rec) == 0))
-				return;
+				return false;
 			rec->flags--;
 

kernel/trace/ftrace.c static void __ftrace_hash_rec_update(struct ftrace_ops *ops, · line 1758

Quoted 7 lines

kernel/trace/ftrace.c +2 -0

 			/*

@@ -1753,22 +1754,27 @@ static void __ftrace_hash_rec_update(struct ftrace_ops *ops,
 			 */
 		}
 		count++;
+
+		update |= ftrace_test_record(rec, 1) != FTRACE_UPDATE_IGNORE;

Shouldn't it use 'inc' instead of 1 for the second argument of the ftrace_test_record()?

On Fri, Mar 11, 2016 at 11:28:00PM +0900, Namhyung Kim wrote:

SNIP

Quoted 17 lines

Quoted 17 lines

kernel/trace/ftrace.c +2 -2

@@ -1694,7 +1695,7 @@ static void __ftrace_hash_rec_update(struct ftrace_ops *ops,
 		if (inc) {
 			rec->flags++;
 			if (FTRACE_WARN_ON(ftrace_rec_count(rec) == FTRACE_REF_MAX))
-				return;
+				return false;
 
 			/*
 			 * If there's only a single callback registered to a

@@ -1720,7 +1721,7 @@ static void __ftrace_hash_rec_update(struct ftrace_ops *ops,
 				rec->flags |= FTRACE_FL_REGS;
 		} else {
 			if (FTRACE_WARN_ON(ftrace_rec_count(rec) == 0))
-				return;
+				return false;
 			rec->flags--;
 

Quoted 10 lines

kernel/trace/ftrace.c static void __ftrace_hash_rec_update(struct ftrace_ops *ops, · line 1758

Quoted 7 lines

kernel/trace/ftrace.c +2 -0

 			/*

@@ -1753,22 +1754,27 @@ static void __ftrace_hash_rec_update(struct ftrace_ops *ops,
 			 */
 		}
 		count++;
+
+		update |= ftrace_test_record(rec, 1) != FTRACE_UPDATE_IGNORE;

Shouldn't it use 'inc' instead of 1 for the second argument of the ftrace_test_record()?

I dont think so, 1 is to update calls (FTRACE_UPDATE_CALLS) check ftrace_modify_all_code:

if (command & FTRACE_UPDATE_CALLS)
ftrace_replace_code(1);
else if (command & FTRACE_DISABLE_CALLS)
ftrace_replace_code(0);

both ftrace_startup, ftrace_shutdown use FTRACE_UPDATE_CALLS

you'd use 0 only to disable all, check ftrace_check_record comments:

        /*
         * If we are updating calls:
         *
         *   If the record has a ref count, then we need to enable it
         *   because someone is using it.
         *
         *   Otherwise we make sure its disabled.
         *
         * If we are disabling calls, then disable all records that
         * are enabled.
         */
        if (enable && ftrace_rec_count(rec))
                flag = FTRACE_FL_ENABLED;

used by ftrace_shutdown_sysctl

Hi Jiri,

Quoted 21 lines · Jiri Olsa

On Fri, Mar 11, 2016 at 11:28:00PM +0900, Namhyung Kim wrote:

SNIP

Quoted 17 lines

Quoted 17 lines

kernel/trace/ftrace.c +2 -2

@@ -1694,7 +1695,7 @@ static void __ftrace_hash_rec_update(struct ftrace_ops *ops,
 		if (inc) {
 			rec->flags++;
 			if (FTRACE_WARN_ON(ftrace_rec_count(rec) == FTRACE_REF_MAX))
-				return;
+				return false;
 
 			/*
 			 * If there's only a single callback registered to a

@@ -1720,7 +1721,7 @@ static void __ftrace_hash_rec_update(struct ftrace_ops *ops,
 				rec->flags |= FTRACE_FL_REGS;
 		} else {
 			if (FTRACE_WARN_ON(ftrace_rec_count(rec) == 0))
-				return;
+				return false;
 			rec->flags--;
 

Quoted 7 lines

Quoted 7 lines

kernel/trace/ftrace.c static void __ftrace_hash_rec_update(struct ftrace_ops *ops, · line 1758

Quoted 7 lines

kernel/trace/ftrace.c +2 -0

 			/*

@@ -1753,22 +1754,27 @@ static void __ftrace_hash_rec_update(struct ftrace_ops *ops,
 			 */
 		}
 		count++;
+
+		update |= ftrace_test_record(rec, 1) != FTRACE_UPDATE_IGNORE;  

Quoted 7 lines

Quoted 3 lines

Shouldn't it use 'inc' instead of 1 for the second argument of the ftrace_test_record()?

I dont think so, 1 is to update calls (FTRACE_UPDATE_CALLS) check ftrace_modify_all_code:

Quoted 6 lines

if (command & FTRACE_UPDATE_CALLS)
ftrace_replace_code(1);
else if (command & FTRACE_DISABLE_CALLS)
ftrace_replace_code(0);

both ftrace_startup, ftrace_shutdown use FTRACE_UPDATE_CALLS

Ah, ok. So the second argument of the ftrace_test_record() is not 'enable' actually.. :-/

Quoted 18 lines

you'd use 0 only to disable all, check ftrace_check_record comments:

        /*
         * If we are updating calls:
         *
         *   If the record has a ref count, then we need to enable it
         *   because someone is using it.
         *
         *   Otherwise we make sure its disabled.
         *
         * If we are disabling calls, then disable all records that
         * are enabled.
         */
        if (enable && ftrace_rec_count(rec))
                flag = FTRACE_FL_ENABLED;

Quoted 1 line

used by ftrace_shutdown_sysctl

I got it. Thank you for the explanation!

Thanks,
Namhyung