Merge tag 'trace-3.8-rc4-fix' of git://git.kernel.org/pub/scm/linux/k…

…ernel/git/rostedt/linux-trace

Pull ftrace fix from Steven Rostedt:
 "Kprobes now uses the function tracer if it can.  That is, if a probe
  is placed on a function mcount/nop location, and the arch supports it,
  instead of adding a breakpoint, kprobes will register a function
  callback as that is much more efficient.

  The function tracer requires to update modules before they run, and
  uses the module notifier to do so.  But if something else in the
  module notifiers registers a kprobe at one of these locations, before
  ftrace can get to it, then the system could fail.

  The function tracer must be initialized early, otherwise module
  notifiers that probe will only work by chance."

* tag 'trace-3.8-rc4-fix' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  ftrace: Be first to run code modification on modules

kernel/trace/ftrace.c

@@ -3998,7 +3998,7 @@ static int ftrace_module_notify(struct notifier_block *self,
 
 struct notifier_block ftrace_module_nb = {
 	.notifier_call = ftrace_module_notify,
-	.priority = 0,
+	.priority = INT_MAX,	/* Run before anything that can use kprobes */
 };
 
 extern unsigned long __start_mcount_loc[];