netdev - Re: [RFC] [PATCH] Don't run __qdisc

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20090729192641.GA3058@ami.dom.local>
Date:	Wed, 29 Jul 2009 21:26:41 +0200
From:	Jarek Poplawski <jarkao2@...il.com>
To:	Krishna Kumar2 <krkumar2@...ibm.com>
Cc:	David Miller <davem@...emloft.net>,
	Herbert Xu <herbert@...dor.apana.org.au>,
	netdev@...r.kernel.org
Subject: Re: [RFC] [PATCH] Don't run __qdisc_run() on a stopped TX queue

On Wed, Jul 29, 2009 at 06:38:36PM +0530, Krishna Kumar2 wrote:
> Hi Jarek,
> 
> Jarek Poplawski <jarkao2@...il.com> wrote on 07/29/2009 06:17:34 PM:
> 
> > Re: [RFC] [PATCH] Don't run __qdisc_run() on a stopped TX queue
> >
> > On Wed, Jul 29, 2009 at 08:30:41PM +0800, Herbert Xu wrote:
> > > On Wed, Jul 29, 2009 at 11:26:14AM +0000, Jarek Poplawski wrote:
> > > >
> > > > If you mean the tx lock there should be no "real" contention: only
> > > > one waiter max. qdisc lock's contention might be higher, but it's
> > > > use (during contention) better: enqueue + dequeue together instead
> > > > of doing it separately.
> > >
> > > Hmm, you will have contention if they're both transmitting a
> > > single flow which must always go into a single physical queue.
> > >
> > > So you'll have two CPUs doing the work of a single CPU, with one
> > > of them always spinning on the TX lock.
> >
> > Hmm.. I'd call it a little waiting, but OK let's call it contention;-)
> > When tx is faster than queue operations there could be no contention
> > at all. I'm not saying I must be right: IMHO it's only worth trying.
> 
> My expectation is that tx would be much longer than a few lines of
> queue operation....

I meant here the case of non-default qdisc, like HTB, HFSC or CBQ,
often with many classes and filters, so a few more lines than usual...

Thanks,
Jarek P.
--
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