get rid of NET_LOCK in the pf purge work

pf purge was moved to systqmp (to get it away from KERNEL_LOCK)
which is also used as the backend for things like intr_barrier and
sched_barrier. it is common for network cards to call intr_barrier
while holding NET_LOCK, and if pf is trying to get the NET_LOCK in
the purge tasks that are now running in systqmp, it's a deadlock.
bluhm@ hit this exact issue.

sashan@ has been working to get rid of the need for NET_LOCK in pf,
so now we can remove the NET_LOCKs here rather than create a pf
specific taskq to run these tasks in.

ok sashan@ bluhm@

1 files changed, 1 insertions, 7 deletions

diff --git a/sys/net/pf.c b/sys/net/pf.c
index 4203911ea60..2552da3ea84 100644
--- a/sys/net/pf.c
+++ b/sys/net/pf.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: pf.c,v 1.1153 2022/11/12 02:48:14 kn Exp $ */
+/*	$OpenBSD: pf.c,v 1.1154 2022/11/25 03:45:39 dlg Exp $ */
 
 /*
  * Copyright (c) 2001 Daniel Hartmeier
@@ -1603,9 +1603,6 @@ pf_purge(void *null)
 {
 	unsigned int interval = max(1, pf_default_rule.timeout[PFTM_INTERVAL]);
 
-	/* XXX is NET_LOCK necessary? */
-	NET_LOCK();
-
 	PF_LOCK();
 
 	pf_purge_expired_src_nodes();
@@ -1616,7 +1613,6 @@ pf_purge(void *null)
 	 * Fragments don't require PF_LOCK(), they use their own lock.
 	 */
 	pf_purge_expired_fragments();
-	NET_UNLOCK();
 
 	/* interpret the interval as idle time between runs */
 	timeout_add_sec(&pf_purge_to, interval);
@@ -1891,7 +1887,6 @@ pf_purge_expired_states(const unsigned int limit, const unsigned int collect)
 	if (SLIST_EMPTY(&gcl))
 		return (scanned);
 
-	NET_LOCK();
 	rw_enter_write(&pf_state_list.pfs_rwl);
 	PF_LOCK();
 	PF_STATE_ENTER_WRITE();
@@ -1904,7 +1899,6 @@ pf_purge_expired_states(const unsigned int limit, const unsigned int collect)
 	PF_STATE_EXIT_WRITE();
 	PF_UNLOCK();
 	rw_exit_write(&pf_state_list.pfs_rwl);
-	NET_UNLOCK();
 
 	while ((st = SLIST_FIRST(&gcl)) != NULL) {
 		SLIST_REMOVE_HEAD(&gcl, gc_list);