ARP Bug?

ARP Bug?

[Matti Linnanvuori]

Jan Engelhardt:

>On Jan 30 2008 04:29, Matti Linnanvuori wrote:
>Jan Engelhardt:
>>> If you have the same subnet on multiple interfaces, only the
>>> first interface will be served.
>>
>>Does that comply with the standard?
>
>What standard?

ARP standard. I think it is RFC 826:
An Ethernet Address Resolution Protocol



      ____________________________________________________________________________________
Never miss a thing.  Make Yahoo your home page. 
http://www.yahoo.com/r/hs

Re: ARP Bug?

[Andi Kleen]

Matti Linnanvuori <mattilinnanvuori@yahoo.com> writes:
>
> ARP standard. I think it is RFC 826:
> An Ethernet Address Resolution Protocol

The correct standard in this case is RFC1122 and it actually
offers two options for this: the so called strong end host model and the 
weak end host model. Linux can be configured to be both. Default
is the later.

-Andi

Re: ARP Bug?

[Jan Engelhardt]

On Jan 30 2008 14:20, Andi Kleen wrote:
>Matti Linnanvuori <mattilinnanvuori@yahoo.com> writes:
>>
>> ARP standard. I think it is RFC 826:
>> An Ethernet Address Resolution Protocol
>
>The correct standard in this case is RFC1122 and it actually
>offers two options for this: the so called strong end host model and the 
>weak end host model. Linux can be configured to be both. Default
>is the later.

By use of a bridge interface, yeah. Are there other ways to get the
strong end host model?

Re: ARP Bug?

[Jan Engelhardt]

On Jan 30 2008 04:44, Matti Linnanvuori wrote:
>>On Jan 30 2008 04:29, Matti Linnanvuori wrote:
>>Jan Engelhardt:
>>>> If you have the same subnet on multiple interfaces, only the
>>>> first interface will be served.
>>>
>>>Does that comply with the standard?
>>
>>What standard?
>
>ARP standard. I think it is RFC 826:
>An Ethernet Address Resolution Protocol

Please quote the section where it says that routing implementations
MUST round-robin over a group of interfaces with same subnet.

ARP Bug?

[Matti Linnanvuori]

Jan Engelhardt:
> If you have the same subnet on multiple interfaces, only the
> first interface will be served.

Does that comply with the standard?




      ____________________________________________________________________________________
Be a better friend, newshound, and 
know-it-all with Yahoo! Mobile.  Try it now.  http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ 

Re: ARP Bug?

[Jan Engelhardt]

On Jan 30 2008 04:29, Matti Linnanvuori wrote:
>Jan Engelhardt:
>> If you have the same subnet on multiple interfaces, only the
>> first interface will be served.
>
>Does that comply with the standard?

What standard?

ARP Bug?

[Gosney, JeremiX]

We've noticed the 2.6-based Linux systems in our test lab are
experiencing some "ARP flux"-like symptoms.  

The systems reply with eth0's hardware address to all ARP requests,
regardless of the IP being queried.  Because of this, the system will
only send and receive packets on eth0; if eth0 is brought down, the
system is unreachable even though it still has several active
connections.  With eth0 unplugged, none of the other interfaces are
reachable (this is presumably a side-effect caused by the switch ARP
cache.) Failover routes are defined in the routing table, but the system
still will not send/receive packets out those interfaces.

Can anything be done to correct this behaviour?


Jeremi M Gosney
Enterprise Linux Engineer
DIJIC, Intel Corporation
Work: 253-371-4849
Mobile: 253-495-4254

Re: ARP Bug?

[Jan Engelhardt]

On Dec 14 2007 17:11, Gosney, JeremiX wrote:
>Subject: ARP Bug?
>
>We've noticed the 2.6-based Linux systems in our test lab are
>experiencing some "ARP flux"-like symptoms.  
>
>The systems reply with eth0's hardware address to all ARP requests,

If you have the same subnet on multiple interfaces, only the
first interface will be served.

Case closed?

>regardless of the IP being queried.  Because of this, the system will
>only send and receive packets on eth0; if eth0 is brought down, the
>system is unreachable even though it still has several active
>connections.  With eth0 unplugged, none of the other interfaces are
>reachable (this is presumably a side-effect caused by the switch ARP
>cache.) Failover routes are defined in the routing table, but the system
>still will not send/receive packets out those interfaces.
>

I am not sure ARP even uses the routing table.

RE: ARP Bug?

[Jan Engelhardt]

On Dec 19 2007 09:47, Peters, Gordon wrote:

>Call me crazy,

For top-posting, yes. :)

> but if I have 2 NICs in a system and all the traffic is
>going through one of them and then that one NIC cable gets pulled, the
>other NIC, that half of the traffic should have been going through in
>the first place, should still work. However, with this type of
>functionality, it does not. Effectively both NICs are now dead even
>though the cable has only been pulled from eth0. Once you pull the cable
>to that "primary" NIC, all traffic stops; even when the "secondary" NIC
>is up and working fine. That in my professional opinion is broken.

I think you should finally reveal your interface configuration
otherwise it's all smoke and mirrors.

Re: ARP Bug?

[Chuck Ebbert]

On 12/14/2007 08:11 PM, Gosney, JeremiX wrote:
> We've noticed the 2.6-based Linux systems in our test lab are
> experiencing some "ARP flux"-like symptoms.  
> 
> The systems reply with eth0's hardware address to all ARP requests,
> regardless of the IP being queried.  Because of this, the system will
> only send and receive packets on eth0; if eth0 is brought down, the
> system is unreachable even though it still has several active
> connections.  With eth0 unplugged, none of the other interfaces are
> reachable (this is presumably a side-effect caused by the switch ARP
> cache.) Failover routes are defined in the routing table, but the system
> still will not send/receive packets out those interfaces.
> 
> Can anything be done to correct this behaviour?
> 
> 

Offtopic here, but anyway look at http://linux-ip.net/html/ch-ether.html