<< | Thread Index | >> ]    [ << | Date Index | >> ]

Subject: Re: QoS / fair queueing in kernel breaks CIPE ...
From: Gert.Vandelaer,AT,medisearch-int,DOT,com
Date: Fri, 31 May 2002 11:41:33 +0200

Hya Mark,

here's how it all went ...

got the 2.4.16 kernel,
make distclean
rolled a new one ...
make bzImage
compiled cipe ; installed it ..
booted new kernel ..
all ok :)

new config kernel
compiled in all QoS options
recompiled cipe too ...
booted new kernel
and from the moment i try ping over VPN I get
"got_short_message"-errors,
and all networking goes down on all interfaces,
plus :
- all routing messages display on the teminal
- everytime unclean dismount of harddrives (ouch)
- finally kernel oops :(

Maybe I need to follow some other procedure,
getting logs is kinda difficult, because after the oops there's not much I
can still do ...

Oh yes, this all happens when there's still no traffic shaping configured
on any interface,
the traffic shaping goes fine when the cipe-module is not loaded and
"ciped-db" daemon hence is not running.

Bye,
Gert

                                                                              
                                     
                    Mark Cooke                                                
                                     
                    <mpc,AT,pc1,DOT,sr,DOT,b        To:     
"Gert.Vandelaer,AT,medisearch-int,DOT,com"                               
                    ham.ac.uk>           
<Gert.Vandelaer,AT,medisearch-int,DOT,com>  
                                     
                                         cc:     "cipe-l,AT,inka,DOT,de" 
<cipe-l,AT,inka,DOT,de>                                 
                    31/05/2002           Subject:     Re: QoS / fair queueing 
in kernel breaks CIPE ...            
                    09:56 AM                                                  
                                     
                                                                              
                                     
                                                                              
                                     

Hi Gert,

I've been using cipe with HTB, and other than a propensity for HTB to log
'HTB delay <number in the range 3000-260000> > 5sec' that appears to have
no obvious detrimental effect, it's been working fine with rate limiting
and traffic classifiction.  I also have the sfq qdisc attached to the cipe
device...

What is your definition of 'screwed up'...

I've been using 2.4.19pre7ac3, and am just compiling 2.4.19pre9ac3.

Mark

My setup script (note I use a HTB patch with my kernel for the eth2
setup to the cable modem):

# QoS setup
tc qdisc add dev eth1 root handle 1: htb default 40
tc class add dev eth1 parent 1: classid 1:1 htb rate 125kbit burst 15k

tc class add dev eth1 parent 1:1 classid 1:10 htb rate 40kbit burst 15k
tc class add dev eth1 parent 1:1 classid 1:20 htb rate 40kbit ceil  125kbit
burst 15k
tc class add dev eth1 parent 1:1 classid 1:30 htb rate 32kbit ceil  100kbit
burst 15k
tc class add dev eth1 parent 1:1 classid 1:40 htb rate 1kbit  ceil  125kbit
burst 15k

tc qdisc add dev eth1 parent 1:10 handle 10: sfq perturb 10
tc qdisc add dev eth1 parent 1:20 handle 20: sfq perturb 10
tc qdisc add dev eth1 parent 1:30 handle 30: sfq perturb 10
tc qdisc add dev eth1 parent 1:40 handle 40: sfq perturb 10

U32="tc filter add dev eth1 protocol ip parent 1:0 prio 1 u32"
# CIPE traffic / ssh
$U32 match ip dport 3434 0xffff flowid 1:10
$U32 match ip dport 3435 0xffff flowid 1:10
$U32 match ip dport 22   0xffff flowid 1:10
# email / http
$U32 match ip dport 80 0xffff flowid 1:20
$U32 match ip dport 25 0xffff flowid 1:20
# netmeeting
$U32 match ip dport 1720 0xffff flowid 1:30

# Add simple fairness
tc qdisc add dev cipc2 root sfq perturb 10

On Fri, 31 May 2002, Gert.Vandelaer,AT,medisearch-int,DOT,com wrote:

> Hi all,
>
> I somebody else would pleast test this also ...
> kernel 2.4.16, cipe latest snapshot (protocol=4)
> the compile goes fine, but when the new kernel is booted
> all the networking is screwed up, not just the cipdb# devices.
>
> Has anyone else noticed this ...
> or maybe there's even something that can be done about this.
>
> I was not trying to apply qdiscs to the cipe device btw ... :)
>
> Bye,
> Gert





<< | Thread Index | >> ]    [ << | Date Index | >> ]