net_sched: move the empty tp check from ->destroy() to ->delete()

We could have a race condition where in ->classify() path we
dereference tp->root and meanwhile a parallel ->destroy() makes it
a NULL. Daniel cured this bug in commit d936377414fa
("net, sched: respect rcu grace period on cls destruction").

This happens when ->destroy() is called for deleting a filter to
check if we are the last one in tp, this tp is still linked and
visible at that time. The root cause of this problem is the semantic
of ->destroy(), it does two things (for non-force case):

1) check if tp is empty
2) if tp is empty we could really destroy it

and its caller, if cares, needs to check its return value to see if it
is really destroyed. Therefore we can't unlink tp unless we know it is
empty.

As suggested by Daniel, we could actually move the test logic to ->delete()
so that we can safely unlink tp after ->delete() tells us the last one is
just deleted and before ->destroy().

Fixes: 1e052be69d04 ("net_sched: destroy proto tp when all filters are gone")
Cc: Roi Dayan <roid@mellanox.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: John Fastabend <john.fastabend@gmail.com>
Cc: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 14 insertions, 13 deletions

diff --git a/net/sched/cls_api.c b/net/sched/cls_api.c
index a8da383b681a..22f88b35a546 100644
--- a/net/sched/cls_api.c
+++ b/net/sched/cls_api.c
@@ -178,14 +178,11 @@ errout:
 	return ERR_PTR(err);
 }
 
-static bool tcf_proto_destroy(struct tcf_proto *tp, bool force)
+static void tcf_proto_destroy(struct tcf_proto *tp)
 {
-	if (tp->ops->destroy(tp, force)) {
-		module_put(tp->ops->owner);
-		kfree_rcu(tp, rcu);
-		return true;
-	}
-	return false;
+	tp->ops->destroy(tp);
+	module_put(tp->ops->owner);
+	kfree_rcu(tp, rcu);
 }
 
 void tcf_destroy_chain(struct tcf_proto __rcu **fl)
@@ -194,7 +191,7 @@ void tcf_destroy_chain(struct tcf_proto __rcu **fl)
 
 	while ((tp = rtnl_dereference(*fl)) != NULL) {
 		RCU_INIT_POINTER(*fl, tp->next);
-		tcf_proto_destroy(tp, true);
+		tcf_proto_destroy(tp);
 	}
 }
 EXPORT_SYMBOL(tcf_destroy_chain);
@@ -361,7 +358,7 @@ replay:
 			RCU_INIT_POINTER(*back, next);
 			tfilter_notify(net, skb, n, tp, fh,
 				       RTM_DELTFILTER, false);
-			tcf_proto_destroy(tp, true);
+			tcf_proto_destroy(tp);
 			err = 0;
 			goto errout;
 		}
@@ -372,24 +369,28 @@ replay:
 			goto errout;
 		}
 	} else {
+		bool last;
+
 		switch (n->nlmsg_type) {
 		case RTM_NEWTFILTER:
 			if (n->nlmsg_flags & NLM_F_EXCL) {
 				if (tp_created)
-					tcf_proto_destroy(tp, true);
+					tcf_proto_destroy(tp);
 				err = -EEXIST;
 				goto errout;
 			}
 			break;
 		case RTM_DELTFILTER:
-			err = tp->ops->delete(tp, fh);
+			err = tp->ops->delete(tp, fh, &last);
 			if (err)
 				goto errout;
 			next = rtnl_dereference(tp->next);
 			tfilter_notify(net, skb, n, tp, t->tcm_handle,
 				       RTM_DELTFILTER, false);
-			if (tcf_proto_destroy(tp, false))
+			if (last) {
 				RCU_INIT_POINTER(*back, next);
+				tcf_proto_destroy(tp);
+			}
 			goto errout;
 		case RTM_GETTFILTER:
 			err = tfilter_notify(net, skb, n, tp, fh,
@@ -411,7 +412,7 @@ replay:
 		tfilter_notify(net, skb, n, tp, fh, RTM_NEWTFILTER, false);
 	} else {
 		if (tp_created)
-			tcf_proto_destroy(tp, true);
+			tcf_proto_destroy(tp);
 	}
 
 errout: