net_sched: fix error path in red_init()
Subsystems (3)
PATCH 1/1 net_sched: fix error path in red_init()
When ->init() fails, ->destroy() is called to clean up. So it is unnecessary to clean up in red_init(), and it would cause some refcount underflow.
--- net/sched/sch_red.c | 20 ++++---------------- 1 file changed, 4 insertions(+), 16 deletions(-)
net/sched/sch_red.c +4 -16
@@ -353,23 +353,11 @@ static int red_init(struct Qdisc *sch, struct nlattr *opt,FLOW_BLOCK_BINDER_TYPE_RED_EARLY_DROP , tb[TCA_RED_EARLY_DROP_BLOCK], extack);if (err) - goto err_early_drop_init; - - err =tcf_qevent_init (&q->qe_mark, sch, -FLOW_BLOCK_BINDER_TYPE_RED_MARK , - tb[TCA_RED_MARK_BLOCK ], extack); -if (err) - goto err_mark_init; - -return 0 ; +return err; -err_mark_init: -tcf_qevent_destroy (&q->qe_early_drop, sch); -err_early_drop_init: -del_timer_sync (&q->adapt_timer); -red_offload (sch, false); -qdisc_put (q->qdisc); -return err; +return tcf_qevent_init (&q->qe_mark, sch, +FLOW_BLOCK_BINDER_TYPE_RED_MARK , + tb[TCA_RED_MARK_BLOCK ], extack); } static int red_change(struct Qdisc *sch, struct nlattr *opt,
Signature · 1 line
2.28.0
Quoted 3 lines · Cong Wang
When ->init() fails, ->destroy() is called to clean up. So it is unnecessary to clean up in
red_init(), and it would cause some refcount underflow.
Hmm, yeah, qdisc_put() would get called twice. A surprising API, the init needs to make sure to always bring the qdisc into destroyable state. But qevents are like that after kzalloc, so the fix looks correct.
Thanks!
From: Cong Wang <[email protected]>
Date: Thu, 27 Aug 2020 10:40:41 -0700
Quoted 9 lines
When ->init() fails, ->destroy() is called to clean up. So it is unnecessary to clean up in
red_init(), and it would cause some refcount underflow.Fixes aee9caa03fc3 ("net: sched: sch_red: Add qevents "early_drop" and "mark"")Reported-and-tested-by [email protected]Reported-and-tested-by [email protected]Cc Petr Machata <[email protected]>Signed-off-by Cong Wang <[email protected]>
Applied, thank you.