[Click] packet weirdness
Roman Chertov
rchertov at purdue.edu
Fri Apr 14 11:54:30 EDT 2006
After reading the doc for FastUDPsrc it seems that it operates on a
single skb_buff. The thing which confuses me is why incr_ports is
allowed to be used in the pull function if packet modifications is not
allowed without cloning.
Roman
Roman Chertov wrote:
> Hello,
> I have a machine with two NICs connected by a cross over cable. I am
> running Click from CVS on a 2.4.32 kernel. I did a few modifications to
> the Click linux module elements such that FastUDPsrc now puts a packet
> sequence number into the packet's payload and then I set the timestamp
> annotation. Then the packet gets sent to eth1. A special element on the
> eth2 recieve path looks at the packets and stores the sequence numbers
> in an internal fixed size array. To get the array data I have made a
> read_handler for it.
>
> If I set the packet generation rate to something low then I get a
> non-repeating set of sequence numbers; however, if I set the rate to 0,
> then I sometimes see a long series of repeating numbers.
>
> Any input would be much appreciated,
>
> Roman
>
> Script.
> source :: FastUDPSource2(0, 10000, 60,
> 00:0E:0C:AB:64:0C, 10.0.1.1, 1234,
> 00:0E:0C:AB:63:70, 10.0.2.1, 1234);
>
> to :: ToDevice(eth1);
> from :: PollDevice(eth2);
> ctr :: AverageCounter;
>
> classifier :: Classifier(12/0800 /* IP packets */,
> - /* everything else */);
> ip_classifier :: IPClassifier(dst udp port 1234 /* relevant UDP
> packets */,
> - /* everything else */);
>
>
> source -> SetTimestamp
> -> to;
>
>
> from -> classifier[0]
> -> CheckIPHeader(14, CHECKSUM false) // don't check checksum for
> speed
> -> ip_classifier[0]
> -> tt :: TimestampTrack(1000)
> -> ctr
> //-> Print
> -> Discard;
>
> classifier[1] -> ToHost;
> ip_classifier[1] -> ToHost;
>
> Idle -> ToDevice(eth2);
> _______________________________________________
> click mailing list
> click at amsterdam.lcs.mit.edu
> https://amsterdam.lcs.mit.edu/mailman/listinfo/click
>
More information about the click
mailing list