[Click] header annotations, variable meanings, etc.

[Eddie Kohler]

That is modifying THE PACKET.  The IP header annotation doesn't contain a COPY 
of the IP header; it contains a POINTER to the ip header.  The IP header data 
itself is part of the packet data, but the POINTER to the header is an annotation.

Eddie


Nicholas Murphy wrote:
> Hmm...ok.  First of all, I don't see any such element in my 
> distribution. :)  But I looked at setipdscp.cc (which is maybe what you 
> were referring to?), and I see the following lines.
> 
>   uint16_t old_hw = (reinterpret_cast<uint16_t *>(ip))[0];
>   ip->ip_tos = (ip->ip_tos & 0x3) | _dscp;
>   uint16_t new_hw = (reinterpret_cast<uint16_t *>(ip))[0];
> 
>   // 19.Aug.1999 - incrementally update IP checksum according to RFC1624.
>   // new_sum = ~(~old_sum + ~old_halfword + new_halfword)
>   uint32_t sum = (~ip->ip_sum & 0xFFFF) + (~old_hw & 0xFFFF) + new_hw;
>   sum = (sum & 0xFFFF) + (sum >> 16);
>   ip->ip_sum = ~(sum + (sum >> 16));
> 
> This seems to be modifying the fields I mentioned before that you said 
> didn't change the original packet.  Will forwarding the packet with 
> these modifications out on a ToDevice not reflect these changes?  I'm 
> confused.
> 
> Sorry for the newbie questions.
> 
> Thanks,
> Nick
> 
> On Dec 10, 2006, at 3:30 PM, Eddie Kohler wrote:
> 
>> Hi Nick,
>>
>> Check out, for example, elements/ip/setipdhcp.cc (the smaction() method).
>>
>> You need to create a WritablePacket using the uniqueify() method, then 
>> you can simply modify its data.
>>
>>    WritablePacket *wp = p->uniqueify();
>>    if (!wp)
>>        // not enough memory to create a new packet
>>        return 0;
>>    wp->data()[0] = 'W';  // or whatever
>>
>> Eddie
>>
>>
>> Nicholas Murphy wrote:
>>> Eddie-
>>> Thanks for your responses.  One quick followup: what is the 
>>> appropriate way to actually make modifications to the packet if 
>>> modifying the annotations doesn't actually change the packet?
>>> Thanks,
>>> Nick
>>> On Dec 9, 2006, at 7:09 PM, Eddie Kohler wrote:
>>>> Hi Nicholas,
>>>>
>>>> Nicholas Murphy wrote:
>>>>> A few random (probably newbie) questions:
>>>>> 1) So, you seem to need to run packets through something like  
>>>>> CheckIPHeader in order to fill out the packet header annotations,  
>>>>> correct?  Do these annotations persist across elements that use  
>>>>> "uniquify", things like Tee, etc.?
>>>>
>>>> Yes.
>>>>
>>>>> 2) Are transport-layer annotations only filled out once you run  
>>>>> through something like IPClassifier?
>>>>
>>>> The CheckIPHeader annotation sets both the start-network-header 
>>>> annotation and the end-network-header==start-transport-header 
>>>> annotation.
>>>>
>>>>> 3) Am I right in thinking that variables like "network_length()"  
>>>>> represent the length from the beginning of the network header to 
>>>>> the  end of the packet?
>>>>
>>>> Yes!
>>>>
>>>>> 4) Am I right in assuming there will never be any layering  
>>>>> information (e.g., ethernet footer) after the data?
>>>>
>>>> In practice yes, although this depends on your network layer.
>>>>
>>>>
>>>>> 5) Does changing the annotation variables (e.g., 
>>>>> ip_header()->ip_len)  actually modify the packet itself?
>>>>
>>>> Nope!
>>>>
>>>> Eddie
>>>>
>>>>> Hopefully these are all easy yes/no's. :)
>>>>> Thanks,
>>>>> Nick
>>>>> _______________________________________________
>>>>> click mailing list
>>>>> click at amsterdam.lcs.mit.edu
>>>>> https://amsterdam.lcs.mit.edu/mailman/listinfo/click