[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1380387822.3596.3.camel@edumazet-glaptop.roam.corp.google.com>
Date: Sat, 28 Sep 2013 10:03:42 -0700
From: Eric Dumazet <eric.dumazet@...il.com>
To: Vijay Subramanian <subramanian.vijay@...il.com>
Cc: netdev@...r.kernel.org, davem@...emloft.net, shemminger@...tta.com,
Mythili Prabhu <mysuryan@...co.com>,
Dave Taht <dave.taht@...ferbloat.net>
Subject: Re: [PATCH v1 net-next] net: pkt_sched: PIE AQM scheme
On Fri, 2013-09-27 at 18:56 -0700, Vijay Subramanian wrote:
> +static int pie_qdisc_enqueue(struct sk_buff *skb, struct Qdisc *sch)
> +{
> + struct pie_sched_data *q = qdisc_priv(sch);
> +
> + if (unlikely(qdisc_qlen(sch) >= sch->limit))
> + goto out;
> +
...
> +out:
> + q->stats.overlimit++;
> + qdisc_drop(skb, sch);
> + return NET_XMIT_CN; /*indicate congestion*/
> +}
If a Qdisc drops a packet because sch->limit is hit, you must :
return qdisc_drop(skb, sch);
So that NET_XMIT_DROP is returned, not NET_XMIT_CN.
This packet was dropped for sure.
vi +96 net/sched/sch_api.c
---enqueue
enqueue returns 0, if packet was enqueued successfully.
If packet (this one or another one) was dropped, it returns
not zero error code.
NET_XMIT_DROP - this packet dropped
Expected action: do not backoff, but wait until queue will clear.
NET_XMIT_CN - probably this packet enqueued, but another one dropped.
Expected action: backoff or ignore
NET_XMIT_POLICED - dropped by police.
Expected action: backoff or error to real-time apps.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists