bpf: decouple the lifetime of cgroup_bpf from cgroup itself

Currently the lifetime of bpf programs attached to a cgroup is bound
to the lifetime of the cgroup itself. It means that if a user
forgets (or intentionally avoids) to detach a bpf program before
removing the cgroup, it will stay attached up to the release of the
cgroup. Since the cgroup can stay in the dying state (the state
between being rmdir()'ed and being released) for a very long time, it
leads to a waste of memory. Also, it blocks a possibility to implement
the memcg-based memory accounting for bpf objects, because a circular
reference dependency will occur. Charged memory pages are pinning the
corresponding memory cgroup, and if the memory cgroup is pinning
the attached bpf program, nothing will be ever released.

A dying cgroup can not contain any processes, so the only chance for
an attached bpf program to be executed is a live socket associated
with the cgroup. So in order to release all bpf data early, let's
count associated sockets using a new percpu refcounter. On cgroup
removal the counter is transitioned to the atomic mode, and as soon
as it reaches 0, all bpf programs are detached.

Because cgroup_bpf_release() can block, it can't be called from
the percpu ref counter callback directly, so instead an asynchronous
work is scheduled.

The reference counter is not socket specific, and can be used for any
other types of programs, which can be executed from a cgroup-bpf hook
outside of the process context, had such a need arise in the future.

Signed-off-by: Roman Gushchin <guro@fb.com>
Cc: jolsa@redhat.com
Signed-off-by: Alexei Starovoitov <ast@kernel.org>

2 files changed, 45 insertions, 7 deletions

diff --git a/kernel/bpf/cgroup.c b/kernel/bpf/cgroup.c
index fcde0f7b2585..d995edbe816d 100644
--- a/kernel/bpf/cgroup.c
+++ b/kernel/bpf/cgroup.c
@@ -22,12 +22,21 @@
 DEFINE_STATIC_KEY_FALSE(cgroup_bpf_enabled_key);
 EXPORT_SYMBOL(cgroup_bpf_enabled_key);
 
+void cgroup_bpf_offline(struct cgroup *cgrp)
+{
+	cgroup_get(cgrp);
+	percpu_ref_kill(&cgrp->bpf.refcnt);
+}
+
 /**
- * cgroup_bpf_put() - put references of all bpf programs
- * @cgrp: the cgroup to modify
+ * cgroup_bpf_release() - put references of all bpf programs and
+ *                        release all cgroup bpf data
+ * @work: work structure embedded into the cgroup to modify
  */
-void cgroup_bpf_put(struct cgroup *cgrp)
+static void cgroup_bpf_release(struct work_struct *work)
 {
+	struct cgroup *cgrp = container_of(work, struct cgroup,
+					   bpf.release_work);
 	enum bpf_cgroup_storage_type stype;
 	unsigned int type;
 
@@ -47,6 +56,22 @@ void cgroup_bpf_put(struct cgroup *cgrp)
 		}
 		bpf_prog_array_free(cgrp->bpf.effective[type]);
 	}
+
+	percpu_ref_exit(&cgrp->bpf.refcnt);
+	cgroup_put(cgrp);
+}
+
+/**
+ * cgroup_bpf_release_fn() - callback used to schedule releasing
+ *                           of bpf cgroup data
+ * @ref: percpu ref counter structure
+ */
+static void cgroup_bpf_release_fn(struct percpu_ref *ref)
+{
+	struct cgroup *cgrp = container_of(ref, struct cgroup, bpf.refcnt);
+
+	INIT_WORK(&cgrp->bpf.release_work, cgroup_bpf_release);
+	queue_work(system_wq, &cgrp->bpf.release_work);
 }
 
 /* count number of elements in the list.
@@ -167,7 +192,12 @@ int cgroup_bpf_inherit(struct cgroup *cgrp)
  */
 #define	NR ARRAY_SIZE(cgrp->bpf.effective)
 	struct bpf_prog_array __rcu *arrays[NR] = {};
-	int i;
+	int ret, i;
+
+	ret = percpu_ref_init(&cgrp->bpf.refcnt, cgroup_bpf_release_fn, 0,
+			      GFP_KERNEL);
+	if (ret)
+		return ret;
 
 	for (i = 0; i < NR; i++)
 		INIT_LIST_HEAD(&cgrp->bpf.progs[i]);
@@ -183,6 +213,9 @@ int cgroup_bpf_inherit(struct cgroup *cgrp)
 cleanup:
 	for (i = 0; i < NR; i++)
 		bpf_prog_array_free(arrays[i]);
+
+	percpu_ref_exit(&cgrp->bpf.refcnt);
+
 	return -ENOMEM;
 }
 
diff --git a/kernel/cgroup/cgroup.c b/kernel/cgroup/cgroup.c
index 217cec4e22c6..ef9cfbfc82a9 100644
--- a/kernel/cgroup/cgroup.c
+++ b/kernel/cgroup/cgroup.c
@@ -4955,8 +4955,6 @@ static void css_release_work_fn(struct work_struct *work)
 		if (cgrp->kn)
 			RCU_INIT_POINTER(*(void __rcu __force **)&cgrp->kn->priv,
 					 NULL);
-
-		cgroup_bpf_put(cgrp);
 	}
 
 	mutex_unlock(&cgroup_mutex);
@@ -5482,6 +5480,8 @@ static int cgroup_destroy_locked(struct cgroup *cgrp)
 
 	cgroup1_check_for_release(parent);
 
+	cgroup_bpf_offline(cgrp);
+
 	/* put the base reference */
 	percpu_ref_kill(&cgrp->self.refcnt);
 
@@ -6221,6 +6221,7 @@ void cgroup_sk_alloc(struct sock_cgroup_data *skcd)
 		 * Don't use cgroup_get_live().
 		 */
 		cgroup_get(sock_cgroup_ptr(skcd));
+		cgroup_bpf_get(sock_cgroup_ptr(skcd));
 		return;
 	}
 
@@ -6232,6 +6233,7 @@ void cgroup_sk_alloc(struct sock_cgroup_data *skcd)
 		cset = task_css_set(current);
 		if (likely(cgroup_tryget(cset->dfl_cgrp))) {
 			skcd->val = (unsigned long)cset->dfl_cgrp;
+			cgroup_bpf_get(cset->dfl_cgrp);
 			break;
 		}
 		cpu_relax();
@@ -6242,7 +6244,10 @@ void cgroup_sk_alloc(struct sock_cgroup_data *skcd)
 
 void cgroup_sk_free(struct sock_cgroup_data *skcd)
 {
-	cgroup_put(sock_cgroup_ptr(skcd));
+	struct cgroup *cgrp = sock_cgroup_ptr(skcd);
+
+	cgroup_bpf_put(cgrp);
+	cgroup_put(cgrp);
 }
 
 #endif	/* CONFIG_SOCK_CGROUP_DATA */