[PATCH v2] cgroups: separate destroy_work into two separate wq

Tadeusz Struk posted 1 patch 1 year, 10 months ago
include/linux/cgroup-defs.h |  5 +++--
kernel/cgroup/cgroup.c      | 14 +++++++-------
2 files changed, 10 insertions(+), 9 deletions(-)
[PATCH v2] cgroups: separate destroy_work into two separate wq
Posted by Tadeusz Struk 1 year, 10 months ago
Syzbot found a corrupted list bug scenario that can be triggered from
cgroup css_create(). The reproduces writes to cgroup.subtree_control
file, which invokes cgroup_apply_control_enable(), css_create(), and
css_populate_dir(), which then randomly fails with a fault injected -ENOMEM.
In such scenario the css_create() error path rcu enqueues css_free_rwork_fn
work for an css->refcnt initialized with css_release() destructor,
and there is a chance that the css_release() function will be invoked
for a cgroup_subsys_state, for which a destroy_work has already been
queued via css_create() error path. This causes a list_add corruption
as can be seen in the syzkaller report [1].
This can be fixed by separating the css_release and ref_kill paths
to work with two separate work_structs.

[1] https://syzkaller.appspot.com/bug?id=e26e54d6eac9d9fb50b221ec3e4627b327465dbd

Cc: Tejun Heo <tj@kernel.org>
Cc: Zefan Li <lizefan.x@bytedance.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Christian Brauner <brauner@kernel.org>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Andrii Nakryiko <andrii@kernel.org>
Cc: Martin KaFai Lau <kafai@fb.com>
Cc: Song Liu <songliubraving@fb.com>
Cc: Yonghong Song <yhs@fb.com>
Cc: John Fastabend <john.fastabend@gmail.com>
Cc: KP Singh <kpsingh@kernel.org>
Cc: <cgroups@vger.kernel.org>
Cc: <netdev@vger.kernel.org>
Cc: <bpf@vger.kernel.org>
Cc: <stable@vger.kernel.org>
Cc: <linux-kernel@vger.kernel.org>

Reported-and-tested-by: syzbot+e42ae441c3b10acf9e9d@syzkaller.appspotmail.com
Fixes: 8f36aaec9c92 ("cgroup: Use rcu_work instead of explicit rcu and work item")
Signed-off-by: Tadeusz Struk <tadeusz.struk@linaro.org>
---
v2: Add a separate work_struct for the css_ref_kill path instead of
    checking if a work has already been enqueued.
---
 include/linux/cgroup-defs.h |  5 +++--
 kernel/cgroup/cgroup.c      | 14 +++++++-------
 2 files changed, 10 insertions(+), 9 deletions(-)

diff --git a/include/linux/cgroup-defs.h b/include/linux/cgroup-defs.h
index 1bfcfb1af352..92b0c5e8c472 100644
--- a/include/linux/cgroup-defs.h
+++ b/include/linux/cgroup-defs.h
@@ -178,8 +178,9 @@ struct cgroup_subsys_state {
 	 */
 	atomic_t online_cnt;
 
-	/* percpu_ref killing and RCU release */
-	struct work_struct destroy_work;
+	/* percpu_ref killing, css release, and RCU release work structs */
+	struct work_struct release_work;
+	struct work_struct killed_ref_work;
 	struct rcu_work destroy_rwork;
 
 	/*
diff --git a/kernel/cgroup/cgroup.c b/kernel/cgroup/cgroup.c
index adb820e98f24..3e00a793e15d 100644
--- a/kernel/cgroup/cgroup.c
+++ b/kernel/cgroup/cgroup.c
@@ -5099,7 +5099,7 @@ static struct cftype cgroup_base_files[] = {
  *    css_free_work_fn().
  *
  * It is actually hairier because both step 2 and 4 require process context
- * and thus involve punting to css->destroy_work adding two additional
+ * and thus involve punting to css->release_work adding two additional
  * steps to the already complex sequence.
  */
 static void css_free_rwork_fn(struct work_struct *work)
@@ -5154,7 +5154,7 @@ static void css_free_rwork_fn(struct work_struct *work)
 static void css_release_work_fn(struct work_struct *work)
 {
 	struct cgroup_subsys_state *css =
-		container_of(work, struct cgroup_subsys_state, destroy_work);
+		container_of(work, struct cgroup_subsys_state, release_work);
 	struct cgroup_subsys *ss = css->ss;
 	struct cgroup *cgrp = css->cgroup;
 
@@ -5210,8 +5210,8 @@ static void css_release(struct percpu_ref *ref)
 	struct cgroup_subsys_state *css =
 		container_of(ref, struct cgroup_subsys_state, refcnt);
 
-	INIT_WORK(&css->destroy_work, css_release_work_fn);
-	queue_work(cgroup_destroy_wq, &css->destroy_work);
+	INIT_WORK(&css->release_work, css_release_work_fn);
+	queue_work(cgroup_destroy_wq, &css->release_work);
 }
 
 static void init_and_link_css(struct cgroup_subsys_state *css,
@@ -5546,7 +5546,7 @@ int cgroup_mkdir(struct kernfs_node *parent_kn, const char *name, umode_t mode)
 static void css_killed_work_fn(struct work_struct *work)
 {
 	struct cgroup_subsys_state *css =
-		container_of(work, struct cgroup_subsys_state, destroy_work);
+		container_of(work, struct cgroup_subsys_state, killed_ref_work);
 
 	mutex_lock(&cgroup_mutex);
 
@@ -5567,8 +5567,8 @@ static void css_killed_ref_fn(struct percpu_ref *ref)
 		container_of(ref, struct cgroup_subsys_state, refcnt);
 
 	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);
+		INIT_WORK(&css->killed_ref_work, css_killed_work_fn);
+		queue_work(cgroup_destroy_wq, &css->killed_ref_work);
 	}
 }
 
-- 
2.36.1