From mboxrd@z Thu Jan 1 00:00:00 1970 From: Dave Taht Subject: Re: netem and hierarchical ingress traffic shaping Date: Fri, 23 Dec 2011 18:54:01 +0100 Message-ID: References: <1324185132.8451.439.camel@denise.theartistscloset.com> <20111218115549.0d9e1259@nehalam.linuxnetplumber.net> <1324661639.4540.18.camel@edumazet-HP-Compaq-6005-Pro-SFF-PC> <1324661950.4540.20.camel@edumazet-HP-Compaq-6005-Pro-SFF-PC> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: Stephen Hemminger , "John A. Sullivan III" , netdev@vger.kernel.org To: Eric Dumazet Return-path: Received: from mail-iy0-f174.google.com ([209.85.210.174]:60470 "EHLO mail-iy0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754224Ab1LWRyC convert rfc822-to-8bit (ORCPT ); Fri, 23 Dec 2011 12:54:02 -0500 Received: by iaeh11 with SMTP id h11so15404118iae.19 for ; Fri, 23 Dec 2011 09:54:01 -0800 (PST) In-Reply-To: <1324661950.4540.20.camel@edumazet-HP-Compaq-6005-Pro-SFF-PC> Sender: netdev-owner@vger.kernel.org List-ID: On Fri, Dec 23, 2011 at 6:39 PM, Eric Dumazet = wrote: > Le vendredi 23 d=E9cembre 2011 =E0 18:33 +0100, Eric Dumazet a =E9cri= t : >> Le dimanche 18 d=E9cembre 2011 =E0 11:55 -0800, Stephen Hemminger a = =E9crit : >> > On Sun, 18 Dec 2011 00:12:12 -0500 >> > "John A. Sullivan III" wrote: >> > >> > > Since netem appears to be classless, we realized we would need t= o >> > > replace the SFQ on each leaf with netem which we really didn't w= ant to >> > > do - not only to not lose SFQ but because we didn't want to main= tain the >> > > netem parameters on each leaf. =A0So, we activated our ifb1 inte= rface, >> > > placed netem on it and redirected all the egress traffic to ifb1= =2E =A0Taht >> > > worked fine. >> > >> > Current versions of netem can take one class. >> > -- >> >> Hmm, I can see that (commit 10f6dfcfde884441) >> >> But it wont work very well, it assumes qdisc uses a single queue if >> netem reordering is requested : >> >> (__skb_queue_head(&q->qdisc->q, skb)) >> >> We should allow reordering if netem queue is changed from tfifo, onl= y if >> new qdisc is compatible with __skb_queue_head(&q->qdisc->q, skb) >> >> (maybe providing a new ->queue_at_head() new ops) > > Also, child qdisc must not scratch skb->cb[], since netem stores > time_to_send in it. Are there any place where all 48 bytes of cb are used? I wouldn't mind if 'time_to_send' became a separate skb field for a more generic 'time_in_queue'... > I guess nobody actually tried this netem mis-feature :( > > I'll send a fix. > > > > -- > To unsubscribe from this list: send the line "unsubscribe netdev" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at =A0http://vger.kernel.org/majordomo-info.html --=20 Dave T=E4ht SKYPE: davetaht US Tel: 1-239-829-5608 =46R Tel: 0638645374 http://www.bufferbloat.net