svn commit: r287849 - stable/10/sys/kern
Konstantin Belousov
kib at FreeBSD.org
Wed Sep 16 04:35:24 UTC 2015
Author: kib
Date: Wed Sep 16 04:35:23 2015
New Revision: 287849
URL: https://svnweb.freebsd.org/changeset/base/287849
Log:
MFC r287366:
Use SLIST_FOREACH_SAFE() to fix iteration.
Modified:
stable/10/sys/kern/kern_event.c
Directory Properties:
stable/10/ (props changed)
Modified: stable/10/sys/kern/kern_event.c
==============================================================================
--- stable/10/sys/kern/kern_event.c Wed Sep 16 04:27:12 2015 (r287848)
+++ stable/10/sys/kern/kern_event.c Wed Sep 16 04:35:23 2015 (r287849)
@@ -1849,7 +1849,7 @@ void
knote(struct knlist *list, long hint, int lockflags)
{
struct kqueue *kq;
- struct knote *kn;
+ struct knote *kn, *tkn;
int error;
if (list == NULL)
@@ -1861,14 +1861,13 @@ knote(struct knlist *list, long hint, in
list->kl_lock(list->kl_lockarg);
/*
- * If we unlock the list lock (and set KN_INFLUX), we can eliminate
- * the kqueue scheduling, but this will introduce four
- * lock/unlock's for each knote to test. If we do, continue to use
- * SLIST_FOREACH, SLIST_FOREACH_SAFE is not safe in our case, it is
- * only safe if you want to remove the current item, which we are
- * not doing.
+ * If we unlock the list lock (and set KN_INFLUX), we can
+ * eliminate the kqueue scheduling, but this will introduce
+ * four lock/unlock's for each knote to test. Also, marker
+ * would be needed to keep iteration position, since filters
+ * or other threads could remove events.
*/
- SLIST_FOREACH(kn, &list->kl_list, kn_selnext) {
+ SLIST_FOREACH_SAFE(kn, &list->kl_list, kn_selnext, tkn) {
kq = kn->kn_kq;
KQ_LOCK(kq);
if ((kn->kn_status & (KN_INFLUX | KN_SCAN)) == KN_INFLUX) {
More information about the svn-src-stable
mailing list