[GIT PULL] scheduler fixes

[Ingo Molnar]

Linus,

Please pull the latest sched-urgent-for-linus git tree from:

   git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git sched-urgent-for-linus

   # HEAD: 8e5bfa8c1f8471aa4a2d30be631ef2b50e10abaf sched/autogroup: Do not use autogroup->tg in zombie threads

Two fixes for autogroup scheduling, for races when turning the feature on/off via 
/proc/sys/kernel/sched_autogroup_enabled.

 Thanks,

	Ingo

------------------>
Oleg Nesterov (2):
      sched/autogroup: Fix autogroup_move_group() to never skip sched_move_task()
      sched/autogroup: Do not use autogroup->tg in zombie threads


 include/linux/sched.h     |  2 ++
 kernel/exit.c             |  1 +
 kernel/sched/auto_group.c | 36 ++++++++++++++++++++++++++++--------
 3 files changed, 31 insertions(+), 8 deletions(-)

diff --git a/include/linux/sched.h b/include/linux/sched.h
index 348f51b0ec92..e9c009dc3a4a 100644
--- a/include/linux/sched.h
+++ b/include/linux/sched.h
@@ -2567,6 +2567,7 @@ extern void sched_autogroup_create_attach(struct task_struct *p);
 extern void sched_autogroup_detach(struct task_struct *p);
 extern void sched_autogroup_fork(struct signal_struct *sig);
 extern void sched_autogroup_exit(struct signal_struct *sig);
+extern void sched_autogroup_exit_task(struct task_struct *p);
 #ifdef CONFIG_PROC_FS
 extern void proc_sched_autogroup_show_task(struct task_struct *p, struct seq_file *m);
 extern int proc_sched_autogroup_set_nice(struct task_struct *p, int nice);
@@ -2576,6 +2577,7 @@ static inline void sched_autogroup_create_attach(struct task_struct *p) { }
 static inline void sched_autogroup_detach(struct task_struct *p) { }
 static inline void sched_autogroup_fork(struct signal_struct *sig) { }
 static inline void sched_autogroup_exit(struct signal_struct *sig) { }
+static inline void sched_autogroup_exit_task(struct task_struct *p) { }
 #endif
 
 extern int yield_to(struct task_struct *p, bool preempt);
diff --git a/kernel/exit.c b/kernel/exit.c
index 9d68c45ebbe3..3076f3089919 100644
--- a/kernel/exit.c
+++ b/kernel/exit.c
@@ -836,6 +836,7 @@ void __noreturn do_exit(long code)
 	 */
 	perf_event_exit_task(tsk);
 
+	sched_autogroup_exit_task(tsk);
 	cgroup_exit(tsk);
 
 	/*
diff --git a/kernel/sched/auto_group.c b/kernel/sched/auto_group.c
index a5d966cb8891..f1c8fd566246 100644
--- a/kernel/sched/auto_group.c
+++ b/kernel/sched/auto_group.c
@@ -111,10 +111,13 @@ bool task_wants_autogroup(struct task_struct *p, struct task_group *tg)
 {
 	if (tg != &root_task_group)
 		return false;
-
 	/*
-	 * We can only assume the task group can't go away on us if
-	 * autogroup_move_group() can see us on ->thread_group list.
+	 * If we race with autogroup_move_group() the caller can use the old
+	 * value of signal->autogroup but in this case sched_move_task() will
+	 * be called again before autogroup_kref_put().
+	 *
+	 * However, there is no way sched_autogroup_exit_task() could tell us
+	 * to avoid autogroup->tg, so we abuse PF_EXITING flag for this case.
 	 */
 	if (p->flags & PF_EXITING)
 		return false;
@@ -122,6 +125,16 @@ bool task_wants_autogroup(struct task_struct *p, struct task_group *tg)
 	return true;
 }
 
+void sched_autogroup_exit_task(struct task_struct *p)
+{
+	/*
+	 * We are going to call exit_notify() and autogroup_move_group() can't
+	 * see this thread after that: we can no longer use signal->autogroup.
+	 * See the PF_EXITING check in task_wants_autogroup().
+	 */
+	sched_move_task(p);
+}
+
 static void
 autogroup_move_group(struct task_struct *p, struct autogroup *ag)
 {
@@ -138,13 +151,20 @@ autogroup_move_group(struct task_struct *p, struct autogroup *ag)
 	}
 
 	p->signal->autogroup = autogroup_kref_get(ag);
-
-	if (!READ_ONCE(sysctl_sched_autogroup_enabled))
-		goto out;
-
+	/*
+	 * We can't avoid sched_move_task() after we changed signal->autogroup,
+	 * this process can already run with task_group() == prev->tg or we can
+	 * race with cgroup code which can read autogroup = prev under rq->lock.
+	 * In the latter case for_each_thread() can not miss a migrating thread,
+	 * cpu_cgroup_attach() must not be possible after cgroup_exit() and it
+	 * can't be removed from thread list, we hold ->siglock.
+	 *
+	 * If an exiting thread was already removed from thread list we rely on
+	 * sched_autogroup_exit_task().
+	 */
 	for_each_thread(p, t)
 		sched_move_task(t);
-out:
+
 	unlock_task_sighand(p, &flags);
 	autogroup_kref_put(prev);
 }