From: mpi <mpi@openbsd.org>
Date: Thu, 25 Jan 2018 14:04:36 +0000 (+0000)
Subject: Assert that ifiq_destroy() is not called with the NET_LOCK() held.
X-Git-Url: http://artulab.com/gitweb/?a=commitdiff_plain;h=043047e1d73eb585ded6355d328b034a624e1e87;p=openbsd

Assert that ifiq_destroy() is not called with the NET_LOCK() held.

Calling taskq_barrier() on a softnet thread while holding the lock
is clearly a deadlock.

ok visa@, dlg@, bluhm@
---

diff --git a/sys/net/ifq.c b/sys/net/ifq.c
index 99b4704f8d0..d790f433a6e 100644
--- a/sys/net/ifq.c
+++ b/sys/net/ifq.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: ifq.c,v 1.21 2018/01/04 11:02:57 tb Exp $ */
+/*	$OpenBSD: ifq.c,v 1.22 2018/01/25 14:04:36 mpi Exp $ */
 
 /*
  * Copyright (c) 2015 David Gwynne <dlg@openbsd.org>
@@ -446,15 +446,8 @@ void
 ifiq_destroy(struct ifiqueue *ifiq)
 {
 	if (!task_del(ifiq->ifiq_softnet, &ifiq->ifiq_task)) {
-		int netlocked = (rw_status(&netlock) == RW_WRITE);
-
-		if (netlocked) /* XXXSMP breaks atomicity */
-			NET_UNLOCK();
-
+		NET_ASSERT_UNLOCKED();
 		taskq_barrier(ifiq->ifiq_softnet);
-
-		if (netlocked)
-			NET_LOCK();
 	}
 
 	/* don't need to lock because this is the last use of the ifiq */