Dummynet and fragments

[Alvaro Saurin]

Hi,

I was wondering if someone could help me with a small problem with  
dummynet. I have a typical dumbbell configuration where I have a  
FreeBSD 6.0 machine with dummynet. It is something like this:

                                                  *----- ubuntu6
                                                  |     10.10.3.6
                      10.10.1.3      10.10.1.4    |
   ubuntu2 ---[hub]--- ubuntu3------- BSD4------[hub]--- ubuntu5
  10.10.2.2     |     10.10.2.3      10.10.3.4          10.10.3.5
                |
   ubuntu1 -----*
  10.10.2.1

I want to set the delay, bandwidth, etc. limit in BSD4, using  
something like

 >        sudo ipfw add 3 pipe 3 all from ubuntu2 to ubuntu6
 >        sudo ipfw add 4 pipe 4 all from ubuntu6 to ubuntu2
 >        sudo ipfw pipe 3 config bw 12000Kbit/s queue 17500bytes  
delay 5ms
 >        sudo ipfw pipe 4 config bw 12000Kbit/s queue 17500bytes  
delay 5ms

for a limit between 'ubuntu2' and 'ubuntu6', leading to a new rule  
set like

 >        sudo ipfw -a list

00003      0        0 pipe 3 ip from 10.10.2.2 to 10.10.3.6
00004      0        0 pipe 4 ip from 10.10.3.6 to 10.10.2.2
00100   1862   144376 allow ip from any to any via lo0
00200      0        0 deny ip from any to 127.0.0.0/8
00300      0        0 deny ip from 127.0.0.0/8 to any
65000 118952 53165334 allow ip from any to any
65100      0        0 deny log logamount 5000 ip from any to any frag
65535      8      512 deny ip from any to any

 >        sudo ipfw pipe show

00003:  12.000 Mbit/s    5 ms  17 KB 0 queues (1 buckets) droptail
00004:  12.000 Mbit/s    5 ms  17 KB 0 queues (1 buckets) droptail

The problem comes here: if I 'ping'  between these two machines,  
everything is fine, but if I 'ping' with a packet size of, ie, 2000,  
no packets arrive at the receiver. Does it have to do with fragmented  
packets? Do I have to include any other rule for dealing with fragments?

And another question: why do I need to specify a 5ms delay for a  
total RTT of 40ms?

Thanks in advance

Alvaro

-- 
Alvaro Saurin <alvaro.saurin at gmail.com> <saurin at dcs.gla.ac.uk>