netdev - Re: Possible regression in HTB

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]

Date:	Thu, 9 Oct 2008 22:18:38 +1100
From:	Simon Horman <horms@...ge.net.au>
To:	Jarek Poplawski <jarkao2@...il.com>
Cc:	Patrick McHardy <kaber@...sh.net>, netdev@...r.kernel.org,
	David Miller <davem@...emloft.net>,
	Martin Devera <devik@....cz>
Subject: Re: Possible regression in HTB

On Thu, Oct 09, 2008 at 06:21:45AM +0000, Jarek Poplawski wrote:
> On Thu, Oct 09, 2008 at 11:54:40AM +1100, Simon Horman wrote:
> > On Wed, Oct 08, 2008 at 08:03:40AM +0000, Jarek Poplawski wrote:
> ...
> > > OK. But as Patrick mentioned it would be interesting to try a little
> > > below hardware limits: 950, or maybe lower, until HTB starts getting
> > > accuracy.
> > 
> > Hi,
> 
> Hi Simon,
> 
> > 
> > it seems that for my particular setup the magic number is 935Mbit/s.

[snip]

> It's hard to believe so small change can matter so much here!!!
> Great testing!

Thanks, I was rather surprised myself.

> It seems it's safer to use something below such magic number, and
> generally 10% below hardware limit could be the rule.

That kind of rule would certainly solve this case,
not it would be nice for such a rule not to be necessary.

> It also looks like my idea that the first class could get not enough
> packets was wrong. It seems it's rather blocked by other classes when
> HTB thinks it can lend more.
> 
> Anyway, I'm also surprised HTB could be still so exact: congratulations
> Martin!
> 
> Simon, if I don't miss something, I guess you should be OK with these
> last changes in requeuing? The older way did some safety for such
> overestimated configs by hiding the real problem, but there were the
> costs: cpu etc.

I think that as Martin pointed out in another post, there
is really too much hypotheses going on at this point.
I think that it would be nice to try and get to the bottom of this,
but I'm prepared to concede that it isn't a show-stopper.

-- 
Simon Horman
  VA Linux Systems Japan K.K., Sydney, Australia Satellite Office
  H: www.vergenet.net/~horms/             W: www.valinux.co.jp/en

--
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