tracing: Place trace_pid_list logic into abstract functions

Instead of having the logic that does trace_pid_list open coded, wrap it in
abstract functions. This will allow a rewrite of the logic that implements
the trace_pid_list without affecting the users.

Note, this causes a change in behavior. Every time a pid is written into
the set_*_pid file, it creates a new list and uses RCU to update it. If
pid_max is lowered, but there was a pid currently in the list that was
higher than pid_max, those pids will now be removed on updating the list.
The old behavior kept that from happening.

The rewrite of the pid_list logic will no longer depend on pid_max,
and will return the old behavior.

Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>

1 files changed, 27 insertions, 51 deletions

diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
index 7896d30d90f7..dcced07a45e6 100644
--- a/kernel/trace/trace.c
+++ b/kernel/trace/trace.c
@@ -512,12 +512,6 @@ int call_filter_check_discard(struct trace_event_call *call, void *rec,
 	return 0;
 }
 
-void trace_free_pid_list(struct trace_pid_list *pid_list)
-{
-	vfree(pid_list->pids);
-	kfree(pid_list);
-}
-
 /**
  * trace_find_filtered_pid - check if a pid exists in a filtered_pid list
  * @filtered_pids: The list of pids to check
@@ -528,14 +522,7 @@ void trace_free_pid_list(struct trace_pid_list *pid_list)
 bool
 trace_find_filtered_pid(struct trace_pid_list *filtered_pids, pid_t search_pid)
 {
-	/*
-	 * If pid_max changed after filtered_pids was created, we
-	 * by default ignore all pids greater than the previous pid_max.
-	 */
-	if (search_pid >= filtered_pids->pid_max)
-		return false;
-
-	return test_bit(search_pid, filtered_pids->pids);
+	return trace_pid_list_is_set(filtered_pids, search_pid);
 }
 
 /**
@@ -592,15 +579,11 @@ void trace_filter_add_remove_task(struct trace_pid_list *pid_list,
 			return;
 	}
 
-	/* Sorry, but we don't support pid_max changing after setting */
-	if (task->pid >= pid_list->pid_max)
-		return;
-
 	/* "self" is set for forks, and NULL for exits */
 	if (self)
-		set_bit(task->pid, pid_list->pids);
+		trace_pid_list_set(pid_list, task->pid);
 	else
-		clear_bit(task->pid, pid_list->pids);
+		trace_pid_list_clear(pid_list, task->pid);
 }
 
 /**
@@ -617,18 +600,19 @@ void trace_filter_add_remove_task(struct trace_pid_list *pid_list,
  */
 void *trace_pid_next(struct trace_pid_list *pid_list, void *v, loff_t *pos)
 {
-	unsigned long pid = (unsigned long)v;
+	long pid = (unsigned long)v;
+	unsigned int next;
 
 	(*pos)++;
 
 	/* pid already is +1 of the actual previous bit */
-	pid = find_next_bit(pid_list->pids, pid_list->pid_max, pid);
+	if (trace_pid_list_next(pid_list, pid, &next) < 0)
+		return NULL;
 
-	/* Return pid + 1 to allow zero to be represented */
-	if (pid < pid_list->pid_max)
-		return (void *)(pid + 1);
+	pid = next;
 
-	return NULL;
+	/* Return pid + 1 to allow zero to be represented */
+	return (void *)(pid + 1);
 }
 
 /**
@@ -645,12 +629,14 @@ void *trace_pid_next(struct trace_pid_list *pid_list, void *v, loff_t *pos)
 void *trace_pid_start(struct trace_pid_list *pid_list, loff_t *pos)
 {
 	unsigned long pid;
+	unsigned int first;
 	loff_t l = 0;
 
-	pid = find_first_bit(pid_list->pids, pid_list->pid_max);
-	if (pid >= pid_list->pid_max)
+	if (trace_pid_list_first(pid_list, &first) < 0)
 		return NULL;
 
+	pid = first;
+
 	/* Return pid + 1 so that zero can be the exit value */
 	for (pid++; pid && l < *pos;
 	     pid = (unsigned long)trace_pid_next(pid_list, (void *)pid, &l))
@@ -686,7 +672,7 @@ int trace_pid_write(struct trace_pid_list *filtered_pids,
 	unsigned long val;
 	int nr_pids = 0;
 	ssize_t read = 0;
-	ssize_t ret = 0;
+	ssize_t ret;
 	loff_t pos;
 	pid_t pid;
 
@@ -699,34 +685,23 @@ int trace_pid_write(struct trace_pid_list *filtered_pids,
 	 * the user. If the operation fails, then the current list is
 	 * not modified.
 	 */
-	pid_list = kmalloc(sizeof(*pid_list), GFP_KERNEL);
+	pid_list = trace_pid_list_alloc();
 	if (!pid_list) {
 		trace_parser_put(&parser);
 		return -ENOMEM;
 	}
 
-	pid_list->pid_max = READ_ONCE(pid_max);
-
-	/* Only truncating will shrink pid_max */
-	if (filtered_pids && filtered_pids->pid_max > pid_list->pid_max)
-		pid_list->pid_max = filtered_pids->pid_max;
-
-	pid_list->pids = vzalloc((pid_list->pid_max + 7) >> 3);
-	if (!pid_list->pids) {
-		trace_parser_put(&parser);
-		kfree(pid_list);
-		return -ENOMEM;
-	}
-
 	if (filtered_pids) {
 		/* copy the current bits to the new max */
-		for_each_set_bit(pid, filtered_pids->pids,
-				 filtered_pids->pid_max) {
-			set_bit(pid, pid_list->pids);
+		ret = trace_pid_list_first(filtered_pids, &pid);
+		while (!ret) {
+			trace_pid_list_set(pid_list, pid);
+			ret = trace_pid_list_next(filtered_pids, pid + 1, &pid);
 			nr_pids++;
 		}
 	}
 
+	ret = 0;
 	while (cnt > 0) {
 
 		pos = 0;
@@ -742,12 +717,13 @@ int trace_pid_write(struct trace_pid_list *filtered_pids,
 		ret = -EINVAL;
 		if (kstrtoul(parser.buffer, 0, &val))
 			break;
-		if (val >= pid_list->pid_max)
-			break;
 
 		pid = (pid_t)val;
 
-		set_bit(pid, pid_list->pids);
+		if (trace_pid_list_set(pid_list, pid) < 0) {
+			ret = -1;
+			break;
+		}
 		nr_pids++;
 
 		trace_parser_clear(&parser);
@@ -756,13 +732,13 @@ int trace_pid_write(struct trace_pid_list *filtered_pids,
 	trace_parser_put(&parser);
 
 	if (ret < 0) {
-		trace_free_pid_list(pid_list);
+		trace_pid_list_free(pid_list);
 		return ret;
 	}
 
 	if (!nr_pids) {
 		/* Cleared the list of pids */
-		trace_free_pid_list(pid_list);
+		trace_pid_list_free(pid_list);
 		read = ret;
 		pid_list = NULL;
 	}