Hello Paolo,
On 1/16/20 1:40 PM, Paolo Abeni wrote:
> I'm sorry for this trial & error experience. I tried to reproduce the
> issue on top of the vcan virtual device, but it looks like it requires
> the timing imposed by a real device, and it's missing here (TL;DR: I
> can't reproduce the issue locally).
No worries. I don't mind testing.
>
> Code wise, the 2nd patch closed a possible race, but it dumbly re-
> opened the one addressed by the first attempt - the 'empty' field must
> be cleared prior to the trylock operation, or we may end-up with such
> field set and the queue not empty.
>
> So, could you please try the following code?
Unfortunately, I still see observe reodering.
Thanks
Ahmad
>
> Many thanks!
> ---
> diff --git a/include/net/pkt_sched.h b/include/net/pkt_sched.h
> index 6a70845bd9ab..fb365fbf65f8 100644
> --- a/include/net/pkt_sched.h
> +++ b/include/net/pkt_sched.h
> @@ -113,7 +113,7 @@ bool sch_direct_xmit(struct sk_buff *skb, struct Qdisc *q,
> struct net_device *dev, struct netdev_queue *txq,
> spinlock_t *root_lock, bool validate);
>
> -void __qdisc_run(struct Qdisc *q);
> +int __qdisc_run(struct Qdisc *q);
>
> static inline void qdisc_run(struct Qdisc *q)
> {
> diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h
> index fceddf89592a..df460fe0773a 100644
> --- a/include/net/sch_generic.h
> +++ b/include/net/sch_generic.h
> @@ -158,7 +158,6 @@ static inline bool qdisc_run_begin(struct Qdisc *qdisc)
> if (qdisc->flags & TCQ_F_NOLOCK) {
> if (!spin_trylock(&qdisc->seqlock))
> return false;
> - WRITE_ONCE(qdisc->empty, false);
> } else if (qdisc_is_running(qdisc)) {
> return false;
> }
> diff --git a/net/core/dev.c b/net/core/dev.c
> index 0ad39c87b7fd..41e89796cc6b 100644
> --- a/net/core/dev.c
> +++ b/net/core/dev.c
> @@ -3624,10 +3624,23 @@ static inline int __dev_xmit_skb(struct sk_buff *skb, struct Qdisc *q,
> end_run:
> qdisc_run_end(q);
> } else {
> + int quota = 0;
> +
> rc = q->enqueue(skb, q, &to_free) & NET_XMIT_MASK;
> - qdisc_run(q);
> + if (READ_ONCE(q->empty))
> + WRITE_ONCE(q->empty, false);
> + if (!qdisc_run_begin(q))
> + goto out;
> +
> + if (likely(!test_bit(__QDISC_STATE_DEACTIVATED,
> + &q->state)))
> + quota = __qdisc_run(q);
> + if (quota > 0)
> + WRITE_ONCE(q->empty, true);
> + qdisc_run_end(q);
> }
>
> +out:
> if (unlikely(to_free))
> kfree_skb_list(to_free);
> return rc;
> diff --git a/net/sched/sch_generic.c b/net/sched/sch_generic.c
> index 5ab696efca95..1bd2c4e9c4c2 100644
> --- a/net/sched/sch_generic.c
> +++ b/net/sched/sch_generic.c
> @@ -376,7 +376,7 @@ static inline bool qdisc_restart(struct Qdisc *q, int *packets)
> return sch_direct_xmit(skb, q, dev, txq, root_lock, validate);
> }
>
> -void __qdisc_run(struct Qdisc *q)
> +int __qdisc_run(struct Qdisc *q)
> {
> int quota = dev_tx_weight;
> int packets;
> @@ -388,6 +388,7 @@ void __qdisc_run(struct Qdisc *q)
> break;
> }
> }
> + return quota;
> }
>
> unsigned long dev_trans_start(struct net_device *dev)
> @@ -649,12 +650,9 @@ static struct sk_buff *pfifo_fast_dequeue(struct Qdisc *qdisc)
>
> skb = __skb_array_consume(q);
> }
> - if (likely(skb)) {
> - qdisc_update_stats_at_dequeue(qdisc, skb);
> - } else {
> - WRITE_ONCE(qdisc->empty, true);
> - }
>
> + if (likely(skb))
> + qdisc_update_stats_at_dequeue(qdisc, skb);
> return skb;
> }
>
>
>
--
Pengutronix e.K. | |
Steuerwalder Str. 21 | http://www.pengutronix.de/ |
31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
