kernel/pid.c: update comment on find_task_by_pid_ns
authorTetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Fri, 5 Mar 2010 21:42:56 +0000 (13:42 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Sat, 6 Mar 2010 19:26:33 +0000 (11:26 -0800)
tasklist_lock does protect the task and its pid, it can't go away.  The
problem is that find_pid_ns() itself is unsafe without rcu lock, it can
race with copy_process()->free_pid(any_pid).

Protecting copy_process()->free_pid(any_pid) with tasklist_lock would make
it possible to call find_task_by_pid_ns() under tasklist safely, but we
don't do so because we are trying to get rid of the read_lock sites of
tasklist_lock.

Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
kernel/pid.c

index b08e697cd83f0944747d5624698d356cf0288090..86b296943e5f274c2b4965a73275661a4c6c727e 100644 (file)
@@ -376,7 +376,7 @@ struct task_struct *pid_task(struct pid *pid, enum pid_type type)
 EXPORT_SYMBOL(pid_task);
 
 /*
- * Must be called under rcu_read_lock() or with tasklist_lock read-held.
+ * Must be called under rcu_read_lock().
  */
 struct task_struct *find_task_by_pid_ns(pid_t nr, struct pid_namespace *ns)
 {