tracing: More clean ups of triggers

From: Steven Rostedt <[email protected]>

In event_trigger_regex_release(), the only code is:

	mutex_lock(&event_mutex);
	if (file->f_mode & FMODE_READ)
		seq_release(inode, file);
	mutex_unlock(&event_mutex);

	return 0;

There's nothing special about the file->f_mode or the seq_release() that requires any locking. Remove the unnecessary locks.

Signed-off-by Steven Rostedt (Google) <[email protected]>

---
 kernel/trace/trace_events_trigger.c | 4 ----
 1 file changed, 4 deletions(-)

kernel/trace/trace_events_trigger.c +0 -4

index 7795af600466..e5dcfcbb2cd5 100644
@@ -314,13 +314,9 @@ static ssize_t event_trigger_regex_write(struct file *file,
 
 static int event_trigger_regex_release(struct inode *inode, struct file *file)
 {
-	mutex_lock(&event_mutex);
-
 	if (file->f_mode & FMODE_READ)
 		seq_release(inode, file);
 
-	mutex_unlock(&event_mutex);
-
 	return 0;
 }
 

Signature · 1 line

2.51.0

From: Steven Rostedt <[email protected]>

The event trigger data requires a full tracepoint_synchronize_unregister() call before freeing. That call can take 100s of milliseconds to complete. In order to allow for bulk freeing of the trigger data, it can not call the tracepoint_synchronize_unregister() for every individual trigger data being free.

Create a kthread that gets created the first time a trigger data is freed, and have it use the lockless llist to get the list of data to free, run the tracepoint_synchronize_unregister() then free everything in the list.

By freeing hundreds of event_trigger_data elements together, it only requires two runs of the synchronization function, and not hundreds of runs. This speeds up the operation by orders of magnitude (milliseconds instead of several seconds).

Signed-off-by Steven Rostedt (Google) <[email protected]>

---
 kernel/trace/trace.h                |  1 +
 kernel/trace/trace_events_trigger.c | 56 +++++++++++++++++++++++++++--
 2 files changed, 54 insertions(+), 3 deletions(-)

kernel/trace/trace.h +1 -0

index 5863800b1ab3..fd5a6daa6c25 100644
@@ -1808,6 +1808,7 @@ struct event_trigger_data {
 	char				*name;
 	struct list_head		named_list;
 	struct event_trigger_data	*named_data;
+	struct llist_node		llist;
 };
 
 /* Avoid typos */

kernel/trace/trace_events_trigger.c +0 -0

index e5dcfcbb2cd5..16e3449f3cfe 100644

kernel/trace/trace_events_trigger.c +33 -2

 #include "trace.h"
 
 static LIST_HEAD(trigger_commands);
 static DEFINE_MUTEX(trigger_cmd_mutex);
 
+static struct task_struct *trigger_kthread;
+static struct llist_head trigger_data_free_list;
+static DEFINE_MUTEX(trigger_data_kthread_mutex);
+
+/* Bulk garbage collection of event_trigger_data elements */
+static int trigger_kthread_fn(void *ignore)
+{
+	struct event_trigger_data *data, *tmp;
+	struct llist_node *llnodes;
+
+	/* Once this task starts, it lives forever */
+	for (;;) {
+		set_current_state(TASK_INTERRUPTIBLE);
+		if (llist_empty(&trigger_data_free_list))
+			schedule();
+
+		__set_current_state(TASK_RUNNING);
+
+		llnodes = llist_del_all(&trigger_data_free_list);
+
+		/* make sure current triggers exit before free */
+		tracepoint_synchronize_unregister();
+
+		llist_for_each_entry_safe(data, tmp, llnodes, llist)
+			kfree(data);
+	}
+
+	return 0;
+}
+
 void trigger_data_free(struct event_trigger_data *data)
 {
 	if (data->cmd_ops->set_filter)
 		data->cmd_ops->set_filter(NULL, data, NULL);
 
-	/* make sure current triggers exit before free */
-	tracepoint_synchronize_unregister();
+	if (unlikely(!trigger_kthread)) {
+		guard(mutex)(&trigger_data_kthread_mutex);
+		/* Check again after taking mutex */

kernel/trace/trace_events_trigger.c +8 -1

+	if (!trigger_kthread) {
+		/* Do it the slow way */
+		tracepoint_synchronize_unregister();
+		kfree(data);
+		return;
+	}
 
-	kfree(data);
+	llist_add(&data->llist, &trigger_data_free_list);
+	wake_up_process(trigger_kthread);
 }
 
 static inline void data_ops_trigger(struct event_trigger_data *data,

Signature · 1 line

2.51.0

Quoted 8 lines

/* Avoid typos */

kernel/trace/trace_events_trigger.c +0 -0

index e5dcfcbb2cd5..16e3449f3cfe 100644
@@ -6,26 +6,76 @@
  */
 

kernel/trace/trace_events_trigger.c line 15

Quoted 8 lines

kernel/trace/trace_events_trigger.c +2 -0

 #include <linux/security.h>
+#include <linux/kthread.h>
 #include <linux/module.h>
 #include <linux/ctype.h>
 #include <linux/mutex.h>
 #include <linux/slab.h>
 #include <linux/rculist.h>
+#include <linux/llist.h>  

nit: Shouldn't we include this in "trace.h" too, because llist_node is used?

Quoted 47 lines

kernel/trace/trace_events_trigger.c +33 -2

 #include "trace.h"
 
 static LIST_HEAD(trigger_commands);
 static DEFINE_MUTEX(trigger_cmd_mutex);
 
+static struct task_struct *trigger_kthread;
+static struct llist_head trigger_data_free_list;
+static DEFINE_MUTEX(trigger_data_kthread_mutex);
+
+/* Bulk garbage collection of event_trigger_data elements */
+static int trigger_kthread_fn(void *ignore)
+{
+	struct event_trigger_data *data, *tmp;
+	struct llist_node *llnodes;
+
+	/* Once this task starts, it lives forever */
+	for (;;) {
+		set_current_state(TASK_INTERRUPTIBLE);
+		if (llist_empty(&trigger_data_free_list))
+			schedule();
+
+		__set_current_state(TASK_RUNNING);
+
+		llnodes = llist_del_all(&trigger_data_free_list);
+
+		/* make sure current triggers exit before free */
+		tracepoint_synchronize_unregister();
+
+		llist_for_each_entry_safe(data, tmp, llnodes, llist)
+			kfree(data);
+	}
+
+	return 0;
+}
+
 void trigger_data_free(struct event_trigger_data *data)
 {
 	if (data->cmd_ops->set_filter)
 		data->cmd_ops->set_filter(NULL, data, NULL);
 
-	/* make sure current triggers exit before free */
-	tracepoint_synchronize_unregister();
+	if (unlikely(!trigger_kthread)) {
+		guard(mutex)(&trigger_data_kthread_mutex);
+		/* Check again after taking mutex */

kernel/trace/trace_events_trigger.c line 69

Quoted 10 lines

kernel/trace/trace_events_trigger.c +10 -0

+		if (!trigger_kthread) {
+			struct task_struct *kthread;
+
+			kthread = kthread_create(trigger_kthread_fn, NULL,
+						 "trigger_data_free");
+			if (!IS_ERR(kthread))
+				WRITE_ONCE(trigger_kthread, kthread);
+		}
+	}
+  

Hmm,

	/* This continues above error case, but we should do it without lock. */

?

Quoted 18 lines

kernel/trace/trace_events_trigger.c +8 -1

+	if (!trigger_kthread) {
+		/* Do it the slow way */
+		tracepoint_synchronize_unregister();
+		kfree(data);
+		return;
+	}
 
-	kfree(data);
+	llist_add(&data->llist, &trigger_data_free_list);
+	wake_up_process(trigger_kthread);
 }
 
 static inline void data_ops_trigger(struct event_trigger_data *data,

Signature · 1 line

2.51.0

From: Steven Rostedt <[email protected]>

The function trigger_process_regex() is called by a few functions, where only one calls strim() on the buffer passed to it. That leaves the other functions not trimming the end of the buffer passed in and making it a little inconsistent.

Remove the strim() from event_trigger_regex_write() and have trigger_process_regex() use strim() instead of skip_spaces(). The buff variable is not passed in as const, so it can be modified.

Signed-off-by Steven Rostedt (Google) <[email protected]>

---
 kernel/trace/trace_events_trigger.c | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

kernel/trace/trace_events_trigger.c +2 -3

index 16e3449f3cfe..1dfe69146a81 100644
@@ -309,7 +309,8 @@ int trigger_process_regex(struct trace_event_file *file, char *buff)
 	char *command, *next;
 	struct event_command *p;
 
-	next = buff = skip_spaces(buff);
+	next = buff = strim(buff);
+
 	command = strsep(&next, ": \t");
 	if (next) {
 		next = skip_spaces(next);

@@ -346,8 +347,6 @@ static ssize_t event_trigger_regex_write(struct file *file,
 	if (IS_ERR(buf))
 		return PTR_ERR(buf);
 
-	strim(buf);
-
 	guard(mutex)(&event_mutex);
 
 	event_file = event_file_file(file);

Signature · 1 line

2.51.0