help on tc with kernels >= 2.6.27

help on tc with kernels >= 2.6.27

[Jorge Bastos]

Hi,
I just joined the mailing list.
I'm using tc to do traffic shaping for about 6 years

Since kernels >= 2.6.27 it just stoped working, and started shaping
traffic to 1/2kb /sec.
How could this be debug? I don't see other place to request for help
other than this one, and I was adviced to ask for help here too.

Thanks in advanced,
Jorge

Re: help on tc with kernels >= 2.6.27

[Marek Kierdelewicz]

>Hi,

Hi Jorge & List,

>I just joined the mailing list.

It's mailing list for netfilter not Linux QoS. I think your issue is
more suitable for lartc mailing list (is lartc list alive by the way?
anyone?).

>I'm using tc to do traffic shaping for about 6 years
>Since kernels >= 2.6.27 it just stoped working, and started shaping
>traffic to 1/2kb /sec.
>How could this be debug? I don't see other place to request for help
>other than this one, and I was adviced to ask for help here too.

Try attaching qdiscs (sfp/pfifo/bfifo) below each htb class. Please
send some more information (tc script, scenario, result of "tc -s -d
class sh dev XXX" command).

Regards,
Marek Kierdelewicz

Re: help on tc with kernels >= 2.6.27

[bsilva]

As of last month at least, the LARTC mailing list is dead.  As near as I
can tell, this is the best list for tc questions, although if I'm wrong,
I'd love to hear about it.  I'm working on a QoS project as well.


Brad


On Fri, 2 Jan 2009, Marek Kierdelewicz wrote:

> >Hi,
>
> Hi Jorge & List,
>
> >I just joined the mailing list.
>
> It's mailing list for netfilter not Linux QoS. I think your issue is
> more suitable for lartc mailing list (is lartc list alive by the way?
> anyone?).
>
> >I'm using tc to do traffic shaping for about 6 years
> >Since kernels >= 2.6.27 it just stoped working, and started shaping
> >traffic to 1/2kb /sec.
> >How could this be debug? I don't see other place to request for help
> >other than this one, and I was adviced to ask for help here too.
>
> Try attaching qdiscs (sfp/pfifo/bfifo) below each htb class. Please
> send some more information (tc script, scenario, result of "tc -s -d
> class sh dev XXX" command).
>
> Regards,
> Marek Kierdelewicz
> --
> To unsubscribe from this list: send the line "unsubscribe netfilter" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>
>

Re: help on tc with kernels >= 2.6.27

[Marek Kierdelewicz]

>My QoS rules (always were the same)
>---
>/sbin/tc qdisc del dev eth0 root
>/sbin/tc qdisc add dev eth0 root handle 1 cbq bandwidth 100Mbit avpkt
>1000 cell 8
>/sbin/tc class change dev eth0 root cbq weight 10Mbit allot 1514

Sorry, but I'm not familiar with cbq. Try the script below:

-----------------------------CUT-----------------------------
TC="/sbin/tc"
DEV="eth0"
$TC qdisc del dev $DEV root
$TC qdisc add dev $DEV root handle 1: htb default 10
$TC class add dev $DEV parent 1: classid 1:1 htb rate 100Mbit
$TC class add dev $DEV parent 1:1 classid 1:10 htb rate 90Mbit ceil \
100Mbit
$TC class add dev $DEV parent 1: classid 1:1280 htb rate \
200kbit ceil 200kbit 
$TC qdisc add dev $DEV parent 1:1280 handle 1280: sfq perturb 1
$TC filter add dev $DEV parent 1:0 protocol ip prio 100 u32 match ip \
src 195.23.14.71 match ip sport 110 0xffff classid 1:1280
-----------------------------CUT-----------------------------

I hope it does the same your cbq.init did - shapes traffic to
195.23.14.71 to maximal throuput of 200kbit/s and lets the rest of the
traffic to go fullspeed.

Regards,
Marek Kierdelewicz

LARTC (was: Re: help on tc with kernels >= 2.6.27)

[/dev/rob0]

On Fri January 2 2009 13:49:08 Marek Kierdelewicz wrote:
> more suitable for lartc mailing list (is lartc list alive by the way?
> anyone?).

I haven't seen anything from LARTC list in about a year.
-- 
    Offlist mail to this address is discarded unless
    "/dev/rob0" or "not-spam" is in Subject: header

Re: help on tc with kernels >= 2.6.27

[Michele Petrazzo - Unipex srl]

Marek Kierdelewicz wrote:
>> Hi,
> 
> Hi Jorge & List,

Hi,

> 
>> I just joined the mailing list.
> 
> It's mailing list for netfilter not Linux QoS. I think your issue is
>  more suitable for lartc mailing list (is lartc list alive by the
> way? anyone?).
> 

If I remember correctly, lartc is dead and it's joined to netfilter one.

>> I'm using tc to do traffic shaping for about 6 years Since kernels
>>> = 2.6.27 it just stoped working, and started shaping traffic to
>> 1/2kb /sec. How could this be debug? I don't see other place to 
>> request for help other than this one, and I was adviced to ask for 
>> help here too.
> 

Just tried here (with 2.6.27) with Marek example and some modifies (like
remove the tc filter and add classify into mangle) and it's works well.

Looking at the .27 changelog at:
http://kernel.org/pub/linux/kernel/v2.6/ChangeLog-2.6.27
seem that there are some patches against the various htb/hfs and so on,
that can (only a supposition) introduce a problems.

Can you recreate a simple working script that show us the problem?

Michele

Re: help on tc with kernels >= 2.6.27

[Michele Petrazzo - Unipex srl]

Jorge Bastos wrote:
> Hi Michele,
> 

Ciao,

> For htb, the script Marek wrote here, shape's the traffic, but it 
> gets a bit unstable, sometimes if goes down, other goes up, for a 
> considered number of seconds, so I believe it's a pike.
> 

I have a "big" environ with htb (kernel .26) and it works well, without
problems. Never seen up and down course.


> Here's what Marek wrote here:
<-cut->

> 
> With cbq, the traffic was always stable at a constant speed, the 
> oscillation wasn't more than about 1kb.
> 

Also here the same.

> What I use for cbq, is the cbq.init script that generates this rules:
> 
> 
> 
> 

<-cut->

> For cbq, instead of the 20kb/sec, it goes to 1/2, sometimes 4 or 5, 
> but never more than that.
> 


With that rules, also here :( . Starting to say that I'm not a big tc
expert but after some test, it's seem that tbf create the problems:
if you try to remove that rule, it'll work. Also tried sfq or cbq like last
queuing and seem to shape correctly. (my test was done with scp and iptraf)
Have you tried it?

Michele

Re: help on tc with kernels >= 2.6.27

[Michele Petrazzo - Unipex srl]

Jorge Bastos wrote:
> Bingo!
> 

:)

> So it seams that the problem is in TBF since 2.6.27 >= Checking the
> 2.6.27 changelog as you said, were made a lot of changes regargind
> TBF (also some CBQ and HTB changes), so that may be the problem. I
> say this because my rules were the same always, and always worked ok.
>  How can this be reported to the QoS people that maintains it in the
> kernel?
> 

See into the changelog, there is the mail addrs of the people who made the
commits and, if you want, to the lkml.

Michele

Re: help on tc with kernels >= 2.6.27

[Michele Petrazzo - Unipex srl]

Jorge Bastos wrote:
> Michele, The NIC in where you are shaping, it's a 100mbit or
> 1000mbit?
> 

My nic are 100mbit (or better, are a 1000mbit with a 100mbit switch) and
I have the problems, so I think that aren't this the problem.

Have you tried to wrote to the devels?

Michele

Re: help on tc with kernels >= 2.6.27

[Michele Petrazzo - Unipex srl]

Jorge Bastos wrote:
> Hum... could it be my NIC's with the "sky2" driver, I mean, a sky2 bug
> causing this?

Could be a lot of things!
Have you tried with another board? A simple rtl8139 can be a good choice

> The devel's... to wich list should I report this Michele?
> 

As said, see to the changelog where there are the mail addrs of the 
devels. Try also on lkml

> To "netdev@vger.kernel.org"?
> 

Can be also this

Michele