vfork: make it killable

Make vfork() killable.

Change do_fork(CLONE_VFORK) to do wait_for_completion_killable().  If it
fails we do not return to the user-mode and never touch the memory shared
with our child.

However, in this case we should clear child->vfork_done before return, we
use task_lock() in do_fork()->wait_for_vfork_done() and
complete_vfork_done() to serialize with each other.

Note: now that we use task_lock() we don't really need completion, we
could turn task->vfork_done into "task_struct *wake_up_me" but this needs
some complications.

NOTE: this and the next patches do not affect in-kernel users of
CLONE_VFORK, kernel threads run with all signals ignored including
SIGKILL/SIGSTOP.

However this is obviously the user-visible change.  Not only a fatal
signal can kill the vforking parent, a sub-thread can do execve or
exit_group() and kill the thread sleeping in vfork().

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

include/linux/sched.h

@@ -2372,7 +2372,7 @@ static inline int thread_group_empty(struct task_struct *p)
  * Protects ->fs, ->files, ->mm, ->group_info, ->comm, keyring
  * subscriptions and synchronises with wait4().  Also used in procfs.  Also
  * pins the final release of task.io_context.  Also protects ->cpuset and
- * ->cgroup.subsys[].
+ * ->cgroup.subsys[]. And ->vfork_done.
  *
  * Nests both inside and outside of read_lock(&tasklist_lock).
  * It must not be nested with write_lock_irq(&tasklist_lock),

kernel/fork.c

@@ -670,10 +670,34 @@ struct mm_struct *mm_access(struct task_struct *task, unsigned int mode)
 
 void complete_vfork_done(struct task_struct *tsk)
 {
-	struct completion *vfork_done = tsk->vfork_done;
+	struct completion *vfork;
 
-	tsk->vfork_done = NULL;
-	complete(vfork_done);
+	task_lock(tsk);
+	vfork = tsk->vfork_done;
+	if (likely(vfork)) {
+		tsk->vfork_done = NULL;
+		complete(vfork);
+	}
+	task_unlock(tsk);
+}
+
+static int wait_for_vfork_done(struct task_struct *child,
+				struct completion *vfork)
+{
+	int killed;
+
+	freezer_do_not_count();
+	killed = wait_for_completion_killable(vfork);
+	freezer_count();
+
+	if (killed) {
+		task_lock(child);
+		child->vfork_done = NULL;
+		task_unlock(child);
+	}
+
+	put_task_struct(child);
+	return killed;
 }
 
 /* Please note the differences between mmput and mm_release.
@@ -717,7 +741,8 @@ void mm_release(struct task_struct *tsk, struct mm_struct *mm)
 	 * If we're exiting normally, clear a user-space tid field if
 	 * requested.  We leave this alone when dying by signal, to leave
 	 * the value intact in a core dump, and to save the unnecessary
-	 * trouble otherwise.  Userland only wants this done for a sys_exit.
+	 * trouble, say, a killed vfork parent shouldn't touch this mm.
+	 * Userland only wants this done for a sys_exit.
 	 */
 	if (tsk->clear_child_tid) {
 		if (!(tsk->flags & PF_SIGNALED) &&
@@ -1551,6 +1576,7 @@ long do_fork(unsigned long clone_flags,
 		if (clone_flags & CLONE_VFORK) {
 			p->vfork_done = &vfork;
 			init_completion(&vfork);
+			get_task_struct(p);
 		}
 
 		/*
@@ -1568,10 +1594,8 @@ long do_fork(unsigned long clone_flags,
 			ptrace_event(trace, nr);
 
 		if (clone_flags & CLONE_VFORK) {
-			freezer_do_not_count();
-			wait_for_completion(&vfork);
-			freezer_count();
-			ptrace_event(PTRACE_EVENT_VFORK_DONE, nr);
+			if (!wait_for_vfork_done(p, &vfork))
+				ptrace_event(PTRACE_EVENT_VFORK_DONE, nr);
 		}
 	} else {
 		nr = PTR_ERR(p);