[PATCH -mm 2/2] is_current_single_threaded: don't use ->mmap_sem

[Oleg Nesterov-3]

is_current_single_threaded() can safely miss a freshly forked CLONE_VM
task, but in this case it must not miss its parent. That is why we take
mm->mmap_sem for writing to make sure a thread/task with the same ->mm
can't pass exit_mm() and disappear.

However we can avoid ->mmap_sem and rely on rcu/barriers:

        - if we do not see the exiting parent on thread/process list
          we see the result of list_del_rcu(), in this case we must
          also see the result of list_add_rcu() which does wmb().

        - if we do see the parent but its ->mm == NULL, we need rmb()
          to make sure we can't miss the child.

Signed-off-by: Oleg Nesterov <oleg@...>
---

 lib/is_single_threaded.c |   10 ++++++----
 1 file changed, 6 insertions(+), 4 deletions(-)

--- WAIT/lib/is_single_threaded.c~ISS_2_RCU 2009-07-09 22:16:29.000000000 +0200
+++ WAIT/lib/is_single_threaded.c 2009-07-09 22:54:41.000000000 +0200
@@ -22,8 +22,6 @@ bool is_current_single_threaded(void)
  struct task_struct *p, *t;
  bool ret;
 
- might_sleep();
-
  if (atomic_read(&task->signal->live) != 1)
  return false;
 
@@ -31,7 +29,6 @@ bool is_current_single_threaded(void)
  return true;
 
  ret = false;
- down_write(&mm->mmap_sem);
  rcu_read_lock();
  for_each_process(p) {
  if (unlikely(p->flags & PF_KTHREAD))
@@ -45,12 +42,17 @@ bool is_current_single_threaded(void)
  goto found;
  if (likely(t->mm))
  break;
+ /*
+ * t->mm == NULL. Make sure next_thread/next_task
+ * will see other CLONE_VM tasks which might be
+ * forked before exiting.
+ */
+ smp_rmb();
  } while_each_thread(p, t);
  }
  ret = true;
 found:
  rcu_read_unlock();
- up_write(&mm->mmap_sem);
 
  return ret;
 }

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

[Oleg Nesterov-3]

On 07/09, Oleg Nesterov wrote:

>
> is_current_single_threaded() can safely miss a freshly forked CLONE_VM
> task, but in this case it must not miss its parent. That is why we take
> mm->mmap_sem for writing to make sure a thread/task with the same ->mm
> can't pass exit_mm() and disappear.
>
> However we can avoid ->mmap_sem and rely on rcu/barriers:
>
> - if we do not see the exiting parent on thread/process list
>  we see the result of list_del_rcu(), in this case we must
>  also see the result of list_add_rcu() which does wmb().
>
> - if we do see the parent but its ->mm == NULL, we need rmb()
>  to make sure we can't miss the child.

To simplify the review, below is the code after the patch.

Oleg.

/*
 * Returns true if the task does not share ->mm with another thread/process.
 */
bool is_current_single_threaded(void)
{
        struct task_struct *task = current;
        struct mm_struct *mm = task->mm;
        struct task_struct *p, *t;
        bool ret;

        if (atomic_read(&task->signal->live) != 1)
                return false;

        if (atomic_read(&mm->mm_users) == 1)
                return true;

        ret = false;
        rcu_read_lock();
        for_each_process(p) {
                if (unlikely(p->flags & PF_KTHREAD))
                        continue;
                if (unlikely(p == task->group_leader))
                        continue;

                t = p;
                do {
                        if (unlikely(t->mm == mm))
                                goto found;
                        if (likely(t->mm))
                                break;
                        /*
                         * t->mm == NULL. Make sure next_thread/next_task
                         * will see other CLONE_VM tasks which might be
                         * forked before exiting.
                         */
                        smp_rmb();
                } while_each_thread(p, t);
        }
        ret = true;
found:
        rcu_read_unlock();

        return ret;
}

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/