[Click] No network header/Segmentation fault
Beyers Cronje
bcronje at gmail.com
Mon Jul 7 05:45:52 EDT 2008
PS you would need to use Unstrip(14) before forwarding to tun if you use the
config below.
Alternatively you can leave out Strip/Unstrip and just specify the offset of
the IP Header directly in MarkIPHeader i.e. MarkIPHeader(14)
On Mon, Jul 7, 2008 at 11:43 AM, Beyers Cronje <bcronje at gmail.com> wrote:
> Hi,
>
> I think you forgot to use Strip or specified the offset to MarkIPHeader in
> your configuration. Also remember to use Classifier to only push proper IP
> packets to MarkIPHeader, i.e. to get rid of ARP packets etc.
> Your config should start with something like this:
>
> FromDevice(eth0) -> Classifier(12/0800) -> Strip(14) -> MarkIPHeader ->
> .......
>
> Beyers
>
> On Mon, Jul 7, 2008 at 10:33 AM, Tushar Soni <omega_marines at yahoo.com>
> wrote:
>
>> Hi,
>>
>> Thanks for your reply. I tried both MarkIPHeader and CheckIPHeader, but
>> when I use them it gives me segmentation fault. I think the
>> reinterpret_cast<const click_ip *>(p->data()) is not working for this
>> particular decoded packet.
>>
>> But I did try something. In my code after I decode I added the following
>> lines:
>>
>> p->push(sizeof(struct click_ether));
>> p->set_mac_header(p->data(), 14); //This sets the mac header as well as
>> the network header
>> click_ip *iph = p->ip_header(); //Now this works and returns the ip header
>> p->set_network_header(20); //This essentially sets the transport header
>> click_udp *udph = p->udp_header(); //This also works now
>>
>> After doing this when I push the packet, then the IPPrint() element (after
>> stripping 14 bytes that i added) is able to print and there are no
>> segmentation faults. So far so good. I send this packet to the tun. But
>> somehow the packet is being dropped or not processed and it never reaches
>> the application layer. I am thinking the initial reason why this packet does
>> not have any network or transport header maybe causing the kernel to drop
>> it. I checked the checksums of both ip and udp header and they seem to be
>> correct. Would you have any ideas why this is happening???
>>
>> Thanks a lot for your help.
>>
>> Best Regards,
>>
>> Tushar
>>
>> Tºüçhé §übt!é
>>
>>
>> ----- Original Message ----
>> From: Beyers Cronje <bcronje at gmail.com>
>> To: Tushar Soni <omega_marines at yahoo.com>
>> Cc: click at pdos.csail.mit.edu
>> Sent: Friday, July 4, 2008 5:27:59 PM
>> Subject: Re: [Click] No network header/Segmentation fault
>>
>> Hi,
>>
>> You have to use either MarkIPHeader of CheckIPHeader elements to set the
>> IP header annotations.
>>
>> Beyers
>>
>> On Fri, Jul 4, 2008 at 5:06 PM, Tushar Soni <omega_marines at yahoo.com>
>> wrote:
>>
>>> Hi everyone,
>>>
>>> I have a scenario where coding some packets at the access point and
>>> decoding the packets at receiver side using click. Now when i decode the
>>> packet and use Print() element, the packet structure and ip header looks ok.
>>> But when i use IPPrint it gives me segmentation fault. When i try to get the
>>> ip header(Packet::ip_header()) or the udp header(Packet::udp_header()) it
>>> returns null (so this is the reason why IPPrint does not work). The
>>> commented code should explain better about what i am doing and the problems
>>> i am facing. I would be glad if someone could help me figure out what i am
>>> doing wrong. Thanks.
>>>
>>> Tushar
>>>
>>> For coding i combine a certain number of packets, from ip header
>>> onwards(ip header included). So when i decode i should have a packet from
>>> ipheader onwards.
>>>
>>> void
>>> Decoder::check_uncoded(Packet *p) //p is the coded packet
>>> {
>>> FoundPacketsList found_pkts_list;// this is typedef of type :
>>> Vector<Packet *>
>>> WritablePacket *wp = p->uniqueify();
>>> xor_header *xor_h_cod = (xor_header *)wp->data();//i have created my
>>> own header
>>> int entries_cod = xor_h_cod->get_entries();
>>> int coded_hlen = xor_header::get_header_length(entries_cod);
>>> for( int i = 0; i < entries_cod; i++)
>>> {
>>> //_ddhq is a type if queue and here i am basically checking if the
>>> ipid of the unocoded packet
>>> //is present in the coded header that i created
>>> int location = _ddhq.is_present(xor_h_cod->get_entry_ipid(i));
>>> if( location >= 0 )
>>> {
>>> //if the id is in the coded header i add it to a list
>>> found_pkts_list.push_back(_ddhq.packet(location));
>>> }
>>> }
>>> if(found_pkts_list.size() > 0)
>>> {
>>>
>>> bool decodable = false;
>>> //this is the condition for decodability
>>> if( entries_cod - found_pkts_list.size() == 1)
>>> {
>>> decodable = true;
>>> }
>>> //remove my header
>>> wp->pull(coded_hlen);
>>> FoundPacketsList::iterator x = found_pkts_list.begin();
>>> //here i combine the packets in the list with the coded packet to
>>> decode it
>>> while( x < found_pkts_list.end())
>>> {
>>> Packet *p_i = (*x);
>>> unsigned char *xored_data = (unsigned char *)wp->data();
>>> unsigned char *data = (unsigned char *)(p_i->ip_header());
>>> int len = p_i->length();
>>> for (int i = 0; i < len; i++) {
>>> xored_data[i] ^= data[i];
>>> }
>>> x++;
>>> }
>>> if(decodable)
>>> {
>>> //wp = wp->uniqueify();
>>> click_chatter("WOOHOO Packet Decoded");
>>> click_chatter("Length: %u", wp->length());
>>>
>>> click_ip *iph = wp->ip_header();
>>> //PROBLEM IS HERE ip_header() returns nothing
>>> if(!iph)
>>> click_chatter("Shitty packet"); // this is always called
>>> because the iph is never created
>>> else
>>> {
>>> click_udp *udph = wp->udp_header();
>>> if(!udph)
>>> click_chatter("UDP header not found");
>>> }
>>> //here i push the packet to the higher layers
>>> //i use a tun element for this and when i run it gives: no
>>> network header as the error
>>> output(1).push(wp);
>>> }
>>>
>>> }
>>> else
>>> {
>>> click_chatter("None of the packets found in uncoded queue.");
>>> _codedq.push(0,p);
>>> }
>>> }
>>>
>>>
>>>
>>>
>>> Tºüçhé §übt!é
>>>
>>>
>>>
>>>
>>>
>>> _______________________________________________
>>> click mailing list
>>> click at amsterdam.lcs.mit.edu
>>> https://amsterdam.lcs.mit.edu/mailman/listinfo/click
>>>
>>
>>
>>
>
More information about the click
mailing list