[PATCH] net_sched: prio: use qdisc_dequeue_peeked

commit 07bd8df5df4369487812bf85a237322ff3569b77
(sch_sfq: fix peek() implementation) changed sfq to use generic
peek helper.

This makes HFSC complain about a non-work-conserving child qdisc, if
prio with sfq child is used within hfsc:

hfsc peeks into prio qdisc, which will then peek into sfq.
returned skb is stashed in sch->gso_skb.

Next, hfsc tries to dequeue from prio, but prio will call sfq dequeue
directly, which may return NULL instead of previously peeked-at skb.

Have prio call qdisc_dequeue_peeked, so sfq->dequeue() is
not called in this case.

Cc: Eric Dumazet <redacted>
Signed-off-by: Florian Westphal <fw@strlen.de>
---
 Eric, does this look correct to you?
 I am not sure if sfq needs fixing instead of this patch here.

diff --git a/net/sched/sch_prio.c b/net/sched/sch_prio.c
index 2a318f2..b5d56a2 100644
--- a/net/sched/sch_prio.c
+++ b/net/sched/sch_prio.c

@@ -112,7 +112,7 @@ static struct sk_buff *prio_dequeue(struct Qdisc *sch)
 
 	for (prio = 0; prio < q->bands; prio++) {
 		struct Qdisc *qdisc = q->queues[prio];
-		struct sk_buff *skb = qdisc->dequeue(qdisc);
+		struct sk_buff *skb = qdisc_dequeue_peeked(qdisc);
 		if (skb) {
 			qdisc_bstats_update(sch, skb);
 			sch->q.qlen--;

-- 
1.7.3.4

Le mardi 09 août 2011 à 14:04 +0200, Florian Westphal a écrit :

Hi Florian

Are you sure this patch is still needed, after commit 
e1738bd9cecc5c867b0e2996470c1ff20f66ba79
(sch_sfq: fix sfq_enqueue())

I am asking before fully reviewing the code, I dont know who
should/shouldnt call qdisc_dequeue_peeked() instead of qdisc->dequeue()

Thanks !

Eric Dumazet [off-list ref] wrote:

quoted

 Eric, does this look correct to you?
 I am not sure if sfq needs fixing instead of this patch here.

diff --git a/net/sched/sch_prio.c b/net/sched/sch_prio.c
index 2a318f2..b5d56a2 100644
--- a/net/sched/sch_prio.c
+++ b/net/sched/sch_prio.c

@@ -112,7 +112,7 @@ static struct sk_buff *prio_dequeue(struct Qdisc *sch)
 
 	for (prio = 0; prio < q->bands; prio++) {
 		struct Qdisc *qdisc = q->queues[prio];
-		struct sk_buff *skb = qdisc->dequeue(qdisc);
+		struct sk_buff *skb = qdisc_dequeue_peeked(qdisc);
 		if (skb) {
 			qdisc_bstats_update(sch, skb);
 			sch->q.qlen--;

Hi Florian

Are you sure this patch is still needed, after commit 
e1738bd9cecc5c867b0e2996470c1ff20f66ba79
(sch_sfq: fix sfq_enqueue())

Yes, I double checked.  This patch is applied, also
I see HFSC complaints even without that code path touched by
"sch_sfq: fix sfq_enqueue" being hit.

From: Florian Westphal <fw@strlen.de>
Date: Tue,  9 Aug 2011 14:04:43 +0200

commit 07bd8df5df4369487812bf85a237322ff3569b77
(sch_sfq: fix peek() implementation) changed sfq to use generic
peek helper.

This makes HFSC complain about a non-work-conserving child qdisc, if
prio with sfq child is used within hfsc:

hfsc peeks into prio qdisc, which will then peek into sfq.
returned skb is stashed in sch->gso_skb.

Next, hfsc tries to dequeue from prio, but prio will call sfq dequeue
directly, which may return NULL instead of previously peeked-at skb.

Have prio call qdisc_dequeue_peeked, so sfq->dequeue() is
not called in this case.

Cc: Eric Dumazet <redacted>
Signed-off-by: Florian Westphal <fw@strlen.de>

Applied and queued up for -stable, thanks!