When I initially examined the function current_cgns_cgroup_from_root(), I was perplexed by its lack of holding cgroup_mutex. However, after Michal explained the reason[0] to me, I realized that it already holds the namespace_sem. I believe this intricacy could also confuse others, so it would be advisable to include an annotation for clarification. After we replace the cgroup_mutex with RCU read lock, if current doesn't hold the namespace_sem, the root cgroup will be NULL. So let's add a WARN_ON_ONCE() for it. [0]. https://lore.kernel.org/bpf/afdnpo3jz2ic2ampud7swd6so5carkilts2mkygcaw67vbw6yh@5b5mncf7qyet Signed-off-by: Yafang Shao <laoar.shao@xxxxxxxxx> Cc: Michal Koutný <mkoutny@xxxxxxxx> --- kernel/cgroup/cgroup.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/kernel/cgroup/cgroup.c b/kernel/cgroup/cgroup.c index cc6a6d9..baba4b1 100644 --- a/kernel/cgroup/cgroup.c +++ b/kernel/cgroup/cgroup.c @@ -1416,6 +1416,11 @@ static inline struct cgroup *__cset_cgroup_from_root(struct css_set *cset, rcu_read_unlock(); + /* + * The namespace_sem is held by current, so the root cgroup can't + * be umounted. Therefore, we can ensure that the res is non-NULL. + */ + WARN_ON_ONCE(!res); return res; } -- 1.8.3.1
