Re: [Nsis-imp] Assigned numbers for version GIMPS -06

[Xiaoming Fu <[e-mail]>]

Good, this reminds me of the programming error we had before. The 
result was found to be that "Set IP_ROUTER_ALERT ON" in our Linux 
2.6 must be used - which represents the case where RAO value is set 
as 0. (This conforms to RFC2113 that only value 0 is allocated for 
current use.) The appearence of a RAO (having value 0) + port number 
identifies a message being a discovery.

Below is our code segment (for IPv4 and v6 in scoutd in our CASP 
implementation):

/*****************************************************************
  Set IP_ROUTER_ALERT option to "on"  in given socket
  returns 0 on success and -1 on failure
******************************************************************/
int r_alert_recv_on(int socket)
{
   int on = TRUE;


   /* insert ipv4/v6 differentiation here and set correctly */

   if (ip_version_get() == 4)
     {

       if ((setsockopt(socket,SOL_IP,IP_ROUTER_ALERT,
		      (char *) &on,sizeof(on))) < 0)
	return -1;
     }
   else
     {
       if ((setsockopt(socket,SOL_IPV6,IP_ROUTER_ALERT,
		      (char *) &on,sizeof(on))) < 0)
	{
	  /*printf("Error: %i\n\n", errno);*/
	  return -1;
	}
     }

   return 0;
}

Xiaoming

Lang, Christopher wrote:
> To be more specific: 
> 
> We have been using standard raw IP sockets with the IP_ROUTER_ALERT option to intercept packets. This method only works on Linux when the RAO value is 0 (see http://lxr.linux.no/source/net/ipv4/ip_options.c#L428). Also we only looked at IPv4 before we stuck to an RAO of 0. I don't know if raw IPv6 sockets work better or not.
> 
> It appears, from comments from Thomas, that it is possible to use any RAO value if the packets are intercepted using iptables and libipq (using a standard unpatched kernel) :-)
> 
> Chris Lang
> 
> 
>>-----Original Message-----
>>From: Lang, Christopher 
>>Sent: 11 July 2005 10:46
>>To: 'Xiaoming Fu'; Hancock, Robert
>>Cc: [e-mail]
>>Subject: RE: [Nsis-imp] Assigned numbers for version GIMPS -06
>>
>>
>>Hi,
>>
>>We have been using Linux as a platform (with a 2.6.7 kernel). 
>>Using non-zero RAO values we intially found that GIMPS 
>>messages were not being picked up by intermediate nodes. 
>>Inspection of the Kernel source revealed that the Linux 
>>kernel ignores all RAO values except 0. 
>>
>>There also does not appear to be support for picking out any 
>>RAO packets in Windows.
>>
>>Which platform have you been using?
>>
>>Chris Lang
>>
>>
>>>-----Original Message-----
>>>From: Xiaoming Fu [mailto:[e-mail]]
>>>Sent: 07 July 2005 17:47
>>>To: Hancock, Robert
>>>Cc: [e-mail]
>>>Subject: Re: [Nsis-imp] Assigned numbers for version GIMPS -06
>>>
>>>
>>>Hancock, Robert wrote:
>>>
>>>>>My point was, to rely on NLI TLV to determine the size of PI (by 
>>>>>minus other fields) may be simple but not efficient for 
>>
>>processing.
>>
>>>>from C.4.3 (the NLI format):
>>>>   
>>
>>+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
>>
>>>>   |   PI-Length   |    IP-TTL     |IP-Ver |        
>>
>>Reserved       |
>>
>>>>   
>>
>>+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
>>
>>>>   PI-Length = the byte length of the Peer-Identity field
>>>>               (note that the Peer-Identity field itself is padded
>>>>               to a whole number of words)
>>>>
>>>>i'm not sure what encoding the PI as a separate object
>>>
>>>would buy you
>>>
>>>>here.
>>>
>>>Sorry, I have missed this field.
>>>
>>>
>>>>but see for example the end of the second paragraph of 
>>
>>3.1, "If the
>>
>>>>routing state does not exist, it may be possible for GIMPS 
>>>
>>>to send a
>>>
>>>>message anyway, with the same encapsulation as used for a Query
>>>>message."
>>>
>>>Well, a GIMPS engine will always do whatever it is supposed to, as
>>>long as the task is part of the 'spec' and the implementation.
>>>
>>>
>>>>>>- the RAO value of 0 is not realistic and in principle would
>>>>>> be NSLPID dependent. but exactly what the values should be
>>>>>> is a question for John's draft - or something like it ;-)
>>>>>
>>>>>That's what I also guessed but there was a question of 
>>
>>being "0". I
>>
>>>>>like John's statement "An NSLPID must be associated with a 
>>>
>>>unique RAO
>>>
>>>>>value." Assuming a preliminary one for basic testing would
>>>
>>>be needed,
>>>
>>>>>possibly using e.g., "diagnostics/ping/traceroute" function.
>>>>
>>>>
>>>>yep - you'll find the same statement about NSLPIDs in the
>>>
>>>GIMPS draft,
>>>
>>>>btw. but the point is, we can't allocate a 'real' ourselves
>>>
>>>(without
>>>
>>>>going to IANA), so we just picked an easy one. Some IP
>>>
>>>stacks seem to
>>>
>>>>have a problem with values other than 0, in our experience.
>>>>
>>>
>>>Good to know of this. Any example platform?
>>>Xiaoming
>>>
>>>_______________________________________________
>>>NSIS-imp mailing list
>>>[e-mail]
>>>https://www1.ietf.org/mailman/listinfo/nsis-imp
>>>
>>
>>_______________________________________________
>>NSIS-imp mailing list
>>[e-mail]
>>https://www1.ietf.org/mailman/listinfo/nsis-imp
>>
> 
> 
> _______________________________________________
> NSIS-imp mailing list
> [e-mail]
> https://www1.ietf.org/mailman/listinfo/nsis-imp

_______________________________________________
NSIS-imp mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/nsis-imp