pidfd: change do_notify_pidfd() to use __wake_up(poll_to_key(EPOLLIN))

rather than wake_up_all(). This way do_notify_pidfd() won't wakeup the
POLLHUP-only waiters which wait for pid_task() == NULL.

TODO:
    - as Christian pointed out, this asks for the new wake_up_all_poll()
      helper, it can already have other users.

    - we can probably discriminate the PIDFD_THREAD and non-PIDFD_THREAD
      waiters, but this needs more work. See
      https://lore.kernel.org/all/[email protected]/

Signed-off-by: Oleg Nesterov <[email protected]>
Link: https://lore.kernel.org/r/[email protected]
Reviewed-by: Tycho Andersen <[email protected]>
Signed-off-by: Christian Brauner <[email protected]>

Author: oleg-nesterov

kernel/signal.c

@@ -2021,11 +2021,12 @@ int send_sigqueue(struct sigqueue *q, struct pid *pid, enum pid_type type)
 
 void do_notify_pidfd(struct task_struct *task)
 {
-	struct pid *pid;
+	struct pid *pid = task_pid(task);
 
 	WARN_ON(task->exit_state == 0);
-	pid = task_pid(task);
-	wake_up_all(&pid->wait_pidfd);
+
+	__wake_up(&pid->wait_pidfd, TASK_NORMAL, 0,
+			poll_to_key(EPOLLIN | EPOLLRDNORM));
 }
 
 /*