query_name_response: Multiple responses received

Robert Dahlem Robert.Dahlem at gmx.net
Tue Aug 15 18:00:41 GMT 2000


Mac,

On Tue, 15 Aug 2000 18:19:16 +1000, Mac wrote:

                             ^^^^^ That's not quite UK, is it? :-)

>>>I've just noticed I'm getting a load of these
>>>query_name_response: Multiple (2) responses received for a query on subnet 
>>>193.62.43.24 for name INFORMATICS<1d>. This response was from IP 193.62.43.39

>>This may come from a box with two interfaces on the same physical network, both 
>>automagically found by its nmbd. The original query was a broadcast coming in on 
>>both interfaces, the other nmbd responded to both of them (in fact the other box 
>>saw two different packets coming in) and your nmbd got a little embarrased. :-)

>I suspect this is indee the case.  The IP address listed is indeed a machine
>with two physical interfaces on the same physical network (and indeed several
>virtual interfaces (IP aliases) as well.
>
>Surely, this is a bug in one or other end of nmbd.

As far as I can see the situation this is not a bug. The other nmbd has no (easy) 
way to check if both incoming packets are in fact the same packet. 

>Since it's interface aware, couldn't it be at least moderately clever about
>handling this situation?

This would require some stateful inspection of incoming packets and probably a lot 
of memory: how many packets would you want nmbd to cache for a busy network? How 
busy is "busy"? 

AFAIR you don't get anything like a sequence number at application level, at least 
you won't ever get one from an UDP packet. How far have two packets to be away 
from each other on the time stream to be considered different?

>It's clearly NOT wrong to have multiple distinct (but interoperating) IP
>subnets on the same physical LAN, so why can't nmbd cope?

It does. It answered with what was requested. :-)

>Or have I got the configuration wrong somewhere?

Well, you may decide to let nmbd run on one interface only, that would stop it 
from answering requests coming in through the other interfaces. But I suspect 
those other interfaces serve a purpose and nmbd would become "invisible" on the 
other interfaces.

You wrote the other interfaces are virtual and from what I understood they are in 
differnet logical subnets. So suppressing nmbd's actions on the other interfaces 
may not be appropriate. 

>>If you're really concerned about it, you might try to play with "interfaces =". 
>>Be careful, as you could stop nmbd answering queries on the "other" interface.

>That's the 'interfaces =' on the nmbd on the dual homed machine, right?

Yes. And perhaps "bind interfaces only" ...

Remember, its not really a problem. If log file size is an issue for you, 
invalidate the offending line in the source and recompile.

Regards,
        Robert

-- 
---------------------------------------------------------------
Robert.Dahlem at gmx.net           Fax +49-69-432647
---------------------------------------------------------------




More information about the samba mailing list