cgroup: Make operations on the cgroup root_list RCU safe

At present, when we perform operations on the cgroup root_list, we must
hold the cgroup_mutex, which is a relatively heavyweight lock. In reality,
we can make operations on this list RCU-safe, eliminating the need to hold
the cgroup_mutex during traversal. Modifications to the list only occur in
the cgroup root setup and destroy paths, which should be infrequent in a
production environment. In contrast, traversal may occur frequently.
Therefore, making it RCU-safe would be beneficial.

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Tejun Heo <tj@kernel.org>

3 files changed, 19 insertions, 8 deletions

diff --git a/include/linux/cgroup-defs.h b/include/linux/cgroup-defs.h
index 4a6b6b77ccb6..4caab0c6b361 100644
--- a/include/linux/cgroup-defs.h
+++ b/include/linux/cgroup-defs.h
@@ -563,6 +563,7 @@ struct cgroup_root {
 
 	/* A list running through the active hierarchies */
 	struct list_head root_list;
+	struct rcu_head rcu;
 
 	/* Hierarchy-specific flags */
 	unsigned int flags;
diff --git a/kernel/cgroup/cgroup-internal.h b/kernel/cgroup/cgroup-internal.h
index c56071f150f2..5e17f01ced9f 100644
--- a/kernel/cgroup/cgroup-internal.h
+++ b/kernel/cgroup/cgroup-internal.h
@@ -170,7 +170,8 @@ extern struct list_head cgroup_roots;
 
 /* iterate across the hierarchies */
 #define for_each_root(root)						\
-	list_for_each_entry((root), &cgroup_roots, root_list)
+	list_for_each_entry_rcu((root), &cgroup_roots, root_list,	\
+				lockdep_is_held(&cgroup_mutex))
 
 /**
  * for_each_subsys - iterate all enabled cgroup subsystems
diff --git a/kernel/cgroup/cgroup.c b/kernel/cgroup/cgroup.c
index 3a436e4f0da1..19784d44d615 100644
--- a/kernel/cgroup/cgroup.c
+++ b/kernel/cgroup/cgroup.c
@@ -1315,7 +1315,7 @@ static void cgroup_exit_root_id(struct cgroup_root *root)
 
 void cgroup_free_root(struct cgroup_root *root)
 {
-	kfree(root);
+	kfree_rcu(root, rcu);
 }
 
 static void cgroup_destroy_root(struct cgroup_root *root)
@@ -1348,7 +1348,7 @@ static void cgroup_destroy_root(struct cgroup_root *root)
 	spin_unlock_irq(&css_set_lock);
 
 	WARN_ON_ONCE(list_empty(&root->root_list));
-	list_del(&root->root_list);
+	list_del_rcu(&root->root_list);
 	cgroup_root_count--;
 
 	if (!have_favordynmods)
@@ -1389,7 +1389,15 @@ static inline struct cgroup *__cset_cgroup_from_root(struct css_set *cset,
 		}
 	}
 
-	BUG_ON(!res_cgroup);
+	/*
+	 * If cgroup_mutex is not held, the cgrp_cset_link will be freed
+	 * before we remove the cgroup root from the root_list. Consequently,
+	 * when accessing a cgroup root, the cset_link may have already been
+	 * freed, resulting in a NULL res_cgroup. However, by holding the
+	 * cgroup_mutex, we ensure that res_cgroup can't be NULL.
+	 * If we don't hold cgroup_mutex in the caller, we must do the NULL
+	 * check.
+	 */
 	return res_cgroup;
 }
 
@@ -1448,7 +1456,6 @@ static struct cgroup *current_cgns_cgroup_dfl(void)
 static struct cgroup *cset_cgroup_from_root(struct css_set *cset,
 					    struct cgroup_root *root)
 {
-	lockdep_assert_held(&cgroup_mutex);
 	lockdep_assert_held(&css_set_lock);
 
 	return __cset_cgroup_from_root(cset, root);
@@ -1456,7 +1463,9 @@ static struct cgroup *cset_cgroup_from_root(struct css_set *cset,
 
 /*
  * Return the cgroup for "task" from the given hierarchy. Must be
- * called with cgroup_mutex and css_set_lock held.
+ * called with css_set_lock held to prevent task's groups from being modified.
+ * Must be called with either cgroup_mutex or rcu read lock to prevent the
+ * cgroup root from being destroyed.
  */
 struct cgroup *task_cgroup_from_root(struct task_struct *task,
 				     struct cgroup_root *root)
@@ -2031,7 +2040,7 @@ void init_cgroup_root(struct cgroup_fs_context *ctx)
 	struct cgroup_root *root = ctx->root;
 	struct cgroup *cgrp = &root->cgrp;
 
-	INIT_LIST_HEAD(&root->root_list);
+	INIT_LIST_HEAD_RCU(&root->root_list);
 	atomic_set(&root->nr_cgrps, 1);
 	cgrp->root = root;
 	init_cgroup_housekeeping(cgrp);
@@ -2114,7 +2123,7 @@ int cgroup_setup_root(struct cgroup_root *root, u16 ss_mask)
 	 * care of subsystems' refcounts, which are explicitly dropped in
 	 * the failure exit path.
 	 */
-	list_add(&root->root_list, &cgroup_roots);
+	list_add_rcu(&root->root_list, &cgroup_roots);
 	cgroup_root_count++;
 
 	/*