1 files changed, 67 insertions, 17 deletions

diff --git a/kernel/exit.c b/kernel/exit.c
index 6a488ad2dce5..5cbc83e83a5d 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -1538,33 +1538,83 @@ static int wait_consider_task(struct wait_opts *wo, int ptrace,
 		return 0;
 	}
 
-	if (likely(!ptrace) && unlikely(task_ptrace(p))) {
+	/* dead body doesn't have much to contribute */
+	if (p->exit_state == EXIT_DEAD)
+		return 0;
+
+	/* slay zombie? */
+	if (p->exit_state == EXIT_ZOMBIE) {
 		/*
-		 * This child is hidden by ptrace.
-		 * We aren't allowed to see it now, but eventually we will.
+		 * A zombie ptracee is only visible to its ptracer.
+		 * Notification and reaping will be cascaded to the real
+		 * parent when the ptracer detaches.
 		 */
-		wo->notask_error = 0;
-		return 0;
-	}
+		if (likely(!ptrace) && unlikely(task_ptrace(p))) {
+			/* it will become visible, clear notask_error */
+			wo->notask_error = 0;
+			return 0;
+		}
 
-	if (p->exit_state == EXIT_DEAD)
-		return 0;
+		/* we don't reap group leaders with subthreads */
+		if (!delay_group_leader(p))
+			return wait_task_zombie(wo, p);
 
-	/*
-	 * We don't reap group leaders with subthreads.
-	 */
-	if (p->exit_state == EXIT_ZOMBIE && !delay_group_leader(p))
-		return wait_task_zombie(wo, p);
+		/*
+		 * Allow access to stopped/continued state via zombie by
+		 * falling through.  Clearing of notask_error is complex.
+		 *
+		 * When !@ptrace:
+		 *
+		 * If WEXITED is set, notask_error should naturally be
+		 * cleared.  If not, subset of WSTOPPED|WCONTINUED is set,
+		 * so, if there are live subthreads, there are events to
+		 * wait for.  If all subthreads are dead, it's still safe
+		 * to clear - this function will be called again in finite
+		 * amount time once all the subthreads are released and
+		 * will then return without clearing.
+		 *
+		 * When @ptrace:
+		 *
+		 * Stopped state is per-task and thus can't change once the
+		 * target task dies.  Only continued and exited can happen.
+		 * Clear notask_error if WCONTINUED | WEXITED.
+		 */
+		if (likely(!ptrace) || (wo->wo_flags & (WCONTINUED | WEXITED)))
+			wo->notask_error = 0;
+	} else {
+		/*
+		 * If @p is ptraced by a task in its real parent's group,
+		 * hide group stop/continued state when looking at @p as
+		 * the real parent; otherwise, a single stop can be
+		 * reported twice as group and ptrace stops.
+		 *
+		 * If a ptracer wants to distinguish the two events for its
+		 * own children, it should create a separate process which
+		 * takes the role of real parent.
+		 */
+		if (likely(!ptrace) && task_ptrace(p) &&
+		    same_thread_group(p->parent, p->real_parent))
+			return 0;
+
+		/*
+		 * @p is alive and it's gonna stop, continue or exit, so
+		 * there always is something to wait for.
+		 */
+		wo->notask_error = 0;
+	}
 
 	/*
-	 * It's stopped or running now, so it might
-	 * later continue, exit, or stop again.
+	 * Wait for stopped.  Depending on @ptrace, different stopped state
+	 * is used and the two don't interact with each other.
 	 */
-	wo->notask_error = 0;
-
 	if (task_stopped_code(p, ptrace))
 		return wait_task_stopped(wo, ptrace, p);
 
+	/*
+	 * Wait for continued.  There's only one continued state and the
+	 * ptracer can consume it which can confuse the real parent.  Don't
+	 * use WCONTINUED from ptracer.  You don't need or want it.
+	 */
 	return wait_task_continued(wo, p);
 }