RE: ARP routing issue

RE: ARP routing issue

[Steve Iribarne]

Hi Jan,

 
-> default gateway is set to 10.0.22.1, on eth0.
-> 
-> Problem is, if I try to ping from another network 
-> (10.216.0.xx) to 10.0.24.xx, i see the following ARP request:
-> 
-> arp who-has 10.0.22.1 tell 10.0.24.xx
->

You see that coming out the eth0 interface??  

If that is the case it is most definately wrong.  Assuming that your
masks are setup properly.  But I haven't worked on the 2.4 kernel for a
long time so I'm not so sure if what you are seeing is a bug that has
been fixed.

-stv

Re: ARP routing issue

[Jan De Luyck]

On Thursday 06 January 2005 17:06, Steve Iribarne wrote:
> Hi Jan,
>
>
> -> default gateway is set to 10.0.22.1, on eth0.
> ->
> -> Problem is, if I try to ping from another network
> -> (10.216.0.xx) to 10.0.24.xx, i see the following ARP request:
> ->
> -> arp who-has 10.0.22.1 tell 10.0.24.xx
> ->
>
> You see that coming out the eth0 interface??
>
> If that is the case it is most definately wrong.  Assuming that your
> masks are setup properly.  But I haven't worked on the 2.4 kernel for a
> long time so I'm not so sure if what you are seeing is a bug that has
> been fixed.

The network information is:
eth0 10.0.22.xxx mask 255.255.255.0
eth1 10.0.24.xxx mask 255.255.255.0

routing:
10.0.22.0 0.0.0.0 255.255.255.0 eth0
10.0.24.0 0.0.0.0 255.255.255.0 eth1
0.0.0.0  10.0.22.1 0.0.0.0  eth0

Jan

-- 
If a man slept by day, he had little time to work.  That was a
satisfying notion to Escargot.
  -- "The Stone Giant", James P. Blaylock

Re: ARP routing issue

[Paul Rolland]

Hello,

Have a look at /proc/sys/net/conf/XXX/arp_filter :

        
arp_filter - BOOLEAN
        1 - Allows you to have multiple network interfaces on the same
        subnet, and have the ARPs for each interface be answered
        based on whether or not the kernel would route a packet from
        the ARP'd IP out that interface (therefore you must use source
        based routing for this to work). In other words it allows control
        of which cards (usually 1) will respond to an arp request.

        0 - (default) The kernel can respond to arp requests with addresses
        from other interfaces. This may seem wrong but it usually makes
        sense, because it increases the chance of successful communication.
        IP addresses are owned by the complete host on Linux, not by
        particular interfaces. Only for more complex setups like load-
        balancing, does this behaviour cause problems.

Regards,
Paul

Paul Rolland, rol(at)as2917.net
ex-AS2917 Network administrator and Peering Coordinator

--

Please no HTML, I'm not a browser - Pas d'HTML, je ne suis pas un navigateur

"Some people dream of success... while others wake up and work hard at it" 

  

> -----Message d'origine-----
> De : linux-kernel-owner@vger.kernel.org 
> [mailto:linux-kernel-owner@vger.kernel.org] De la part de Jan De Luyck
> Envoyé : jeudi 6 janvier 2005 17:12
> À : Steve Iribarne
> Cc : linux-kernel@vger.kernel.org; linux-net@vger.kernel.org
> Objet : Re: ARP routing issue
> 
> On Thursday 06 January 2005 17:06, Steve Iribarne wrote:
> > Hi Jan,
> >
> >
> > -> default gateway is set to 10.0.22.1, on eth0.
> > ->
> > -> Problem is, if I try to ping from another network
> > -> (10.216.0.xx) to 10.0.24.xx, i see the following ARP request:
> > ->
> > -> arp who-has 10.0.22.1 tell 10.0.24.xx
> > ->
> >
> > You see that coming out the eth0 interface??
> >
> > If that is the case it is most definately wrong.  Assuming that your
> > masks are setup properly.  But I haven't worked on the 2.4 
> kernel for a
> > long time so I'm not so sure if what you are seeing is a 
> bug that has
> > been fixed.
> 
> The network information is:
> eth0 10.0.22.xxx mask 255.255.255.0
> eth1 10.0.24.xxx mask 255.255.255.0
> 
> routing:
> 10.0.22.0 0.0.0.0 255.255.255.0 eth0
> 10.0.24.0 0.0.0.0 255.255.255.0 eth1
> 0.0.0.0  10.0.22.1 0.0.0.0  eth0
> 
> Jan
> 
> -- 
> If a man slept by day, he had little time to work.  That was a
> satisfying notion to Escargot.
>   -- "The Stone Giant", James P. Blaylock
> -
> To unsubscribe from this list: send the line "unsubscribe 
> linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 

Re: ARP routing issue

[Jan De Luyck]

On Thursday 06 January 2005 18:53, Paul Rolland wrote:
> Hello,
>
> Have a look at /proc/sys/net/conf/XXX/arp_filter :
>
>
> arp_filter - BOOLEAN
>         1 - Allows you to have multiple network interfaces on the same
>         subnet, and have the ARPs for each interface be answered
>         based on whether or not the kernel would route a packet from
>         the ARP'd IP out that interface (therefore you must use source
>         based routing for this to work). In other words it allows control
>         of which cards (usually 1) will respond to an arp request.
>
>         0 - (default) The kernel can respond to arp requests with addresses
>         from other interfaces. This may seem wrong but it usually makes
>         sense, because it increases the chance of successful communication.
>         IP addresses are owned by the complete host on Linux, not by
>         particular interfaces. Only for more complex setups like load-
>         balancing, does this behaviour cause problems.
>
> Regards,
> Paul

I tried that actually, didn't change a thing.

Jan

-- 
Beware of computerized fortune-tellers!

Re: ARP routing issue

[James Courtier-Dutton]

Jan De Luyck wrote:
> On Thursday 06 January 2005 17:06, Steve Iribarne wrote:
> 
>>Hi Jan,
>>
>>
>>-> default gateway is set to 10.0.22.1, on eth0.
>>->
>>-> Problem is, if I try to ping from another network
>>-> (10.216.0.xx) to 10.0.24.xx, i see the following ARP request:
>>->
>>-> arp who-has 10.0.22.1 tell 10.0.24.xx
>>->
>>
>>You see that coming out the eth0 interface??
>>
>>If that is the case it is most definately wrong.  Assuming that your
>>masks are setup properly.  But I haven't worked on the 2.4 kernel for a
>>long time so I'm not so sure if what you are seeing is a bug that has
>>been fixed.
> 
> 
> The network information is:
> eth0 10.0.22.xxx mask 255.255.255.0
> eth1 10.0.24.xxx mask 255.255.255.0
> 
> routing:
> 10.0.22.0 0.0.0.0 255.255.255.0 eth0
> 10.0.24.0 0.0.0.0 255.255.255.0 eth1
> 0.0.0.0  10.0.22.1 0.0.0.0  eth0
> 
> Jan
> 

That arp is perfectly OK.
The routing table will cause the icmp echo packet to go from 10.216.0.xx 
to 10.0.24.xx via the 10.0.24.x network.
The icmp echo response will return via the 10.0.22.x network back to the 
10.216.0.xx network.
So the paths in each direction are different.

the "arp who-has 10.0.22.1 tell 10.0.24.xx", you can safely ignore the 
"10.0.24.xx" bit, as that will be ignored by the device responding to 
the ARP.
It is just highlighting the point that you have 2 paths to the same 
destination.

Cheers

James

Re: ARP routing issue

[Jan De Luyck]

On Friday 14 January 2005 23:47, James Courtier-Dutton wrote:
> That arp is perfectly OK.
> The routing table will cause the icmp echo packet to go from 10.216.0.xx
> to 10.0.24.xx via the 10.0.24.x network.
> The icmp echo response will return via the 10.0.22.x network back to the
> 10.216.0.xx network.
> So the paths in each direction are different.

Yes, but unfortunately I never ever receive the icmp echo reply, and the arp 
table always lists the ip as "incomplete". Nothing I try to do to with that 
interface (ssh/...) ever works.

Jan

-- 
Real Programmers don't eat quiche.  They eat Twinkies and Szechwan food.

Re: ARP routing issue

[Alan Cox]

On Sad, 2005-01-15 at 12:31, Jan De Luyck wrote:
> On Friday 14 January 2005 23:47, James Courtier-Dutton wrote:
> > That arp is perfectly OK.
> > The routing table will cause the icmp echo packet to go from 10.216.0.xx
> > to 10.0.24.xx via the 10.0.24.x network.
> > The icmp echo response will return via the 10.0.22.x network back to the
> > 10.216.0.xx network.
> > So the paths in each direction are different.
> 
> Yes, but unfortunately I never ever receive the icmp echo reply, and the arp 
> table always lists the ip as "incomplete". Nothing I try to do to with that 
> interface (ssh/...) ever works.

If the directions are different does your distro enable rp_filter by
default - that may cause such problems. You might also want to ask on
netdev@oss.sgi.com - the network layer list

Re: ARP routing issue - semi-solved

[Jan De Luyck]

Hello lists,

Just for the archive records: we solved it this way (with the good help of IBM):

# ip route add 10.0.24.0/24 dev eth1 proto kernel scope link src 10.0.24.xxx table 1
# ip route add default via 10.0.24.1 dev eth1 table 1 
# ip rule add from 10.0.24.xxx/32 table 1 priority 500
# ip route flush cache

Which is now run at bootup on the affected servers, giving us the solve we need.

Thanks everyone who replied.

Jan
-- 
Auribus teneo lupum.
 [I hold a wolf by the ears.]
 [Boy, it *sounds* good.  But what does it *mean*?]