From: Yafang Shao <laoar.shao@xxxxxxxxx>
commit d23b5c577715892c87533b13923306acc6243f93 upstream.
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@xxxxxxxxx>
Signed-off-by: Tejun Heo <tj@xxxxxxxxxx>
[fp: adapt to 5.10 mainly because of changes made by e210a89f5b07
("cgroup.c: add helper __cset_cgroup_from_root to cleanup duplicated
codes")]
Signed-off-by: Fedor Pchelkin <pchelkin@xxxxxxxxx>
[Shivani: Modified to apply on v4.19.y]
Signed-off-by: Shivani Agarwal <shivani.agarwal@xxxxxxxxxxxx>
---
include/linux/cgroup-defs.h | 1 +
kernel/cgroup/cgroup-internal.h | 2 +-
kernel/cgroup/cgroup.c | 23 ++++++++++++++++-------
3 files changed, 18 insertions(+), 8 deletions(-)
diff --git a/include/linux/cgroup-defs.h b/include/linux/cgroup-defs.h
index 56442d3b651d..1803c222e204 100644
--- a/include/linux/cgroup-defs.h
+++ b/include/linux/cgroup-defs.h
@@ -478,6 +478,7 @@ 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 4168e7d97e87..b96bbbc4b19c 100644
--- a/kernel/cgroup/cgroup-internal.h
+++ b/kernel/cgroup/cgroup-internal.h
@@ -151,7 +151,7 @@ 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)
/**
* for_each_subsys - iterate all enabled cgroup subsystems
diff --git a/kernel/cgroup/cgroup.c b/kernel/cgroup/cgroup.c
index 30c058806702..39f5c00cca29 100644
--- a/kernel/cgroup/cgroup.c
+++ b/kernel/cgroup/cgroup.c
@@ -1246,7 +1246,7 @@ void cgroup_free_root(struct cgroup_root *root)
{
if (root) {
idr_destroy(&root->cgroup_idr);
- kfree(root);
+ kfree_rcu(root, rcu);
}
}
@@ -1280,7 +1280,7 @@ static void cgroup_destroy_root(struct cgroup_root *root)
spin_unlock_irq(&css_set_lock);
if (!list_empty(&root->root_list)) {
- list_del(&root->root_list);
+ list_del_rcu(&root->root_list);
cgroup_root_count--;
}
@@ -1333,7 +1333,6 @@ static struct cgroup *cset_cgroup_from_root(struct css_set *cset,
{
struct cgroup *res = NULL;
- lockdep_assert_held(&cgroup_mutex);
lockdep_assert_held(&css_set_lock);
if (cset == &init_css_set) {
@@ -1353,13 +1352,23 @@ static struct cgroup *cset_cgroup_from_root(struct css_set *cset,
}
}
- BUG_ON(!res);
+ /*
+ * 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;
}
/*
* 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)
@@ -1922,7 +1931,7 @@ void init_cgroup_root(struct cgroup_root *root, struct cgroup_sb_opts *opts)
{
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);
@@ -2004,7 +2013,7 @@ int cgroup_setup_root(struct cgroup_root *root, u16 ss_mask, int ref_flags)
* 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++;
/*
--
2.39.4
