diff --git a/include/linux/cgroup-defs.h b/include/linux/cgroup-defs.h
index 1bfcfb1af3524f..8f9620cbc4a7c6 100644
--- a/include/linux/cgroup-defs.h
+++ b/include/linux/cgroup-defs.h
@@ -182,6 +182,9 @@ struct cgroup_subsys_state {
 	struct work_struct destroy_work;
 	struct rcu_work destroy_rwork;
 
+	/* To serialize kill and release paths for the css */
+	struct completion complete;
+
 	/*
 	 * PI: the parent css.	Placed here for cache proximity to following
 	 * fields of the containing structure.
diff --git a/kernel/cgroup/cgroup.c b/kernel/cgroup/cgroup.c
index 1779ccddb734d0..841f23b4f66904 100644
--- a/kernel/cgroup/cgroup.c
+++ b/kernel/cgroup/cgroup.c
@@ -1592,6 +1592,13 @@ void cgroup_kn_unlock(struct kernfs_node *kn)
 	mutex_unlock(&cgroup_mutex);
 
 	kernfs_unbreak_active_protection(kn);
+	/* If css_kill started for this css wait for it to finish */
+	if (cgrp->self.flags & CSS_DYING) {
+printk("!!!! cgroup_kn_unlock WAITING for comp !!! \n");
+		WARN(!wait_for_completion_timeout(&cgrp->self.complete, msecs_to_jiffies(1000)),
+		     "Timed out waiting for css_killed_ref_fn to finish");
+	}
+
 	cgroup_put(cgrp);
 }
 
@@ -1992,7 +1999,8 @@ int cgroup_setup_root(struct cgroup_root *root, u16 ss_mask)
 	int i, ret;
 
 	lockdep_assert_held(&cgroup_mutex);
-
+printk("!!! cgroup_setup_root init_comp %llx !!! \n", &root_cgrp->self);
+	init_completion(&root_cgrp->self.complete);
 	ret = percpu_ref_init(&root_cgrp->self.refcnt, css_release,
 			      0, GFP_KERNEL);
 	if (ret)
@@ -5109,6 +5117,7 @@ static void css_free_rwork_fn(struct work_struct *work)
 	struct cgroup_subsys *ss = css->ss;
 	struct cgroup *cgrp = css->cgroup;
 
+printk("!!!!!!!!! css_free_rwork_fn css = %llx !!!!!!\n", css);
 	percpu_ref_exit(&css->refcnt);
 
 	if (ss) {
@@ -5210,6 +5219,7 @@ static void css_release(struct percpu_ref *ref)
 	struct cgroup_subsys_state *css =
 		container_of(ref, struct cgroup_subsys_state, refcnt);
 
+printk("!!!! css_release css = %llx !!!\n", css);
 	INIT_WORK(&css->destroy_work, css_release_work_fn);
 	queue_work(cgroup_destroy_wq, &css->destroy_work);
 }
@@ -5307,6 +5317,8 @@ static struct cgroup_subsys_state *css_create(struct cgroup *cgrp,
 	if (IS_ERR(css))
 		return css;
 
+printk("!!! css_create init_comp %llx !!! \n", css);
+	init_completion(&css->complete);
 	init_and_link_css(css, ss, cgrp);
 
 	err = percpu_ref_init(&css->refcnt, css_release, 0, GFP_KERNEL);
@@ -5331,6 +5343,7 @@ static struct cgroup_subsys_state *css_create(struct cgroup *cgrp,
 err_list_del:
 	list_del_rcu(&css->sibling);
 err_free_css:
+printk("!!! css_create ERROR init_comp %llx !!! \n", css);
 	list_del_rcu(&css->rstat_css_node);
 	INIT_RCU_WORK(&css->destroy_rwork, css_free_rwork_fn);
 	queue_rcu_work(cgroup_destroy_wq, &css->destroy_rwork);
@@ -5373,6 +5386,8 @@ static struct cgroup *cgroup_create(struct cgroup *parent, const char *name,
 	}
 	cgrp->kn = kn;
 
+printk("!!! cgroup_create init_comp %llx !!! \n", &cgrp->self);
+	init_completion(&cgrp->self.complete);
 	init_cgroup_housekeeping(cgrp);
 
 	cgrp->self.parent = &parent->self;
@@ -5549,6 +5564,13 @@ static void css_killed_work_fn(struct work_struct *work)
 		container_of(work, struct cgroup_subsys_state, destroy_work);
 
 	mutex_lock(&cgroup_mutex);
+printk("!!!! css_killed_work_fn css = %llx comp %llx !!!\n", css, &css->complete);
+
+	if (!(css->flags & CSS_RELEASED)) {
+		complete(&css->complete);
+	} else {
+printk("!!!! css_killed_work_fn css = %llx comp %llx ALREADY RELEASED !!!\n", css, &css->complete);
+	}
 
 	do {
 		offline_css(css);
@@ -5566,9 +5588,12 @@ static void css_killed_ref_fn(struct percpu_ref *ref)
 	struct cgroup_subsys_state *css =
 		container_of(ref, struct cgroup_subsys_state, refcnt);
 
+printk("!!!! css_killed_ref_fn css = %llx !!!\n", css);
 	if (atomic_dec_and_test(&css->online_cnt)) {
 		INIT_WORK(&css->destroy_work, css_killed_work_fn);
 		queue_work(cgroup_destroy_wq, &css->destroy_work);
+	} else {
+		complete(&css->complete);
 	}
 }
 
@@ -5588,6 +5613,8 @@ static void kill_css(struct cgroup_subsys_state *css)
 	if (css->flags & CSS_DYING)
 		return;
 
+printk("!!! kill_css reinit_comp %llx !!! \n", css);
+	init_completion(&css->complete);
 	css->flags |= CSS_DYING;
 
 	/*
@@ -5772,6 +5799,8 @@ static void __init cgroup_init_subsys(struct cgroup_subsys *ss, bool early)
 	 * newly registered, all tasks and hence the
 	 * init_css_set is in the subsystem's root cgroup. */
 	init_css_set.subsys[ss->id] = css;
+printk("!!! cgroup_init_subsys init_comp %llx !!! \n", css);
+	init_completion(&css->complete);
 
 	have_fork_callback |= (bool)ss->fork << ss->id;
 	have_exit_callback |= (bool)ss->exit << ss->id;
@@ -5875,6 +5904,8 @@ int __init cgroup_init(void)
 			cgroup_init_subsys(ss, false);
 		}
 
+printk("!!! cgroup_init init_comp %llx !!! \n", &init_css_set.subsys[ssid]);
+		init_completion(&init_css_set.subsys[ssid]->complete);
 		list_add_tail(&init_css_set.e_cset_node[ssid],
 			      &cgrp_dfl_root.cgrp.e_csets[ssid]);
 
diff --git a/kernel/sched/completion.c b/kernel/sched/completion.c
index 35f15c26ed54d4..7f1854f6658722 100644
--- a/kernel/sched/completion.c
+++ b/kernel/sched/completion.c
@@ -28,7 +28,6 @@
 void complete(struct completion *x)
 {
 	unsigned long flags;
-
 	raw_spin_lock_irqsave(&x->wait.lock, flags);
 
 	if (x->done != UINT_MAX)