[Xorp-users] Using Xorp for PIM SSM

[pcaldes]

Bruce Simpson wrote:
> Hi,
>
> Thanks for your question regarding XORP's PIM-SM implementation.
>
> pcaldes wrote:
>> ...
>> We figure we should be able to run Xorp on one of the servers and it 
>> can convert IGMPv3 to PIM
>> for other servers on the VLAN in addition to an IGMPv3 application on 
>> the same box
>>   
>
> I believe that should work. I should point out that XORP does not 
> currently support IGMP proxy mode functionality, so if you intend to 
> use XORP as the multicast topology intermediate hop, running PIM-SM is 
> the only option supported.
>
> ...
>>                                             
>> The Xorp router only needs to support SSM so no RP should be needed 
>> but we cannot get it to work.
>> Hopefully it is just a configuration error.
>>   
>
> I believe this is a limitation of the configuration options for XORP's 
> PIM-SM implementation. Not knowing the inner workings in detail, I 
> can't say for sure -- I believe it assumes engineers will be deploying 
> both ASM and SSM models, therefore an RP is required.
>
> A static RP configuration should be fine, providing it doesn't clash 
> with an existing RP in the PIM-SM domain.
>
>> We know our applications send out IGMPv3 SSM joins since if we stop Xorp
>> and enable IGMPv3 on the Cisco router, everything works.
>>
>> Also, is there a configuration command similar to Cisco's "ip pim ssm 
>> range" statement which allows
>> you to specify non-standard SSM subnet ranges?
>>
>>   
>
> Unfortunately no, I believe this is a limitation of the current 
> implementation for much the same reasons.  I am sure this 
> functionality can be added given xorp_pim's decent C++ abstractions.
>
>> Before we start XORP, we make sure to run:
>>
>>     echo 0 > /proc/sys/net/ipv4/conf/eth0/rp_filter
>>   
>
> Just as an aside: I see this is recommended in the XORP User Manual.
>
> I believe this is conceptually similar to FreeBSD's 
> net.inet.ip.check_interface feature, however I believe Linux doesn't 
> fully implement the RFC 1122 Strong End-Station model for the same 
> software stack limitations as FreeBSD has.   I'd be curious to know why.
>  
>>
>>
>> Can you please look over my config file for any mistakes?
>> All help is appreciated.
>>   
>
> Your configuration looks OK, apart from the fact that there is no 
> cand-rp or bootstrap blocks in the pim configuration block.
>
> I believe that the quickest way to resolve your issue may be to just 
> add a cand-rp block to the XORP configuration with a scope-zone for 
> the SSM range you intend to use, even though SSM actually requires no 
> RP, unlike the ASM model.
>
> thanks,
> BMS
>
In our initial configuration we had added the following bootstrap config 
from a sample multicast config we found:
        bootstrap {
            disable: false
            cand-bsr {
                scope-zone 224.0.0.0/4 {
                    cand-bsr-by-vif-name: "eth0"
                }
            }
            cand-rp {
                group-prefix 224.0.0.0/4 {
                    cand-rp-by-vif-name: "eth0"
                }
            }
        }

However, this made us become the RP for other multicast traffic in the 
network which caused some trouble in our network and headaches for us.
The problem is that we don't know in advance what the SSM groups we need 
to join. It is dynamic.
While we could change the group-prefix under cand-rp{} to 232.0.0.0/8 to 
limit us to the default SSM range,
I do not know feel I am knowledgeable about the ramifications of doing 
that since it might still affect other multicast traffic in the cloud.
And I don't want to have to explain ... again .... why I interrupted our 
customer traffic.... again.


Hopefully someone who knows more about the Xorp SSM internals can let us 
know if there is another workaround or how difficult it might be to fix.

Do you think I should repost my original email to the xorp-hackers list? 
I assumed they also read xorp-users and didn't want to crosspost.