[ECOS] RE : [ECOS] RE : [ECOS] Duplicate ARP table entries

["Arnaud Chataignier" <achataignier at neotion dot com>]

Concerning the gateway, I also use a 0.0.0.0 value.

Concerning the direction of ARP requests, the problem disappear if I set
up manually a static ARP entry on host side, meaning that the host
doesn't send an ARP request anymore.

I suspect the problem to appear only when the target sends an ARP
request just a short while after having received an ARP request from the
host (and having answered to it). This happens in the case of a (e.g.
ping) command sent by the host, if the host had no more the IP/ARP
association entry of the target in its ARP table.

Host -------- Target
   ARP req ->
<- ARP ans
   ping req ->
<- ARP req
   ARP ans ->  *** duplicate ARP entry on target ?
<- ping ans    problem : never sent

I investigated the bug for a while two months ago, and I arrived to the
following conclusion : the problem doesn't appear if the execution of
the above flow is executed slowly (I added some RS232 traces at each
step for this). This problem is thus quite hard to investigate and
debug, but it's for sure time-related.

Regards,
Arnaud.

-----Message d'origine-----
De : ecos-discuss-owner@ecos.sourceware.org
[mailto:ecos-discuss-owner@ecos.sourceware.org] De la part de Jeff
Duncan
Envoyé : mercredi 10 novembre 2004 02:54
À : Gary Thomas; Arnaud Chataignier
Cc : ecos-discuss@sources.redhat.com
Objet : RE: [ECOS] RE : [ECOS] Duplicate ARP table entries


More info -

All my ip info is static right now and I'm using the build_bootp
function.  If I set the gateway ip to the same value as the host ip then
there are no problems.

If I set the gateway to a different IP on the same subnet or if I set
the gateway to 0.0.0.0 so that no gateway shows in the routing table
then I get into the situation I described earlier.

Jeff


-----Original Message-----
From: Jeff Duncan 
Sent: Tuesday, November 09, 2004 12:46 PM
To: 'Gary Thomas'; Arnaud Chataignier
Cc: ecos-discuss@sources.redhat.com
Subject: RE: [ECOS] RE : [ECOS] Duplicate ARP table entries


Actually we see this problem on units with a non-Ethernet version of
Redboot as well as units where we pull the MAC from the same location
(no fconfig involved with our modified ethernet driver.)

Also the problem can occur after it's been working just fine and then
sits idle for 20+ minutes.  The ARP entries time out after 20 minutes so
it seems related to ARPs being generated from the device and the host
such that I end up with 2 duplicate entries in the table - one that gets
updated by ARP responses and the other that gets used by ether_output.

Jeff
 

-----Original Message-----
From: Gary Thomas [mailto:gary@mlbassoc.com] 
Sent: Tuesday, November 09, 2004 11:09 AM
To: Arnaud Chataignier
Cc: Jeff Duncan; ecos-discuss@sources.redhat.com
Subject: Re: [ECOS] RE : [ECOS] Duplicate ARP table entries


On Tue, 2004-11-09 at 10:14, Arnaud Chataignier wrote:
> I confirm the bug. I see it also frequently, but unfortunately I 
> hadn't had time to dig into the problem yet.
> 
> What I do as a temporary patch for the moment is to set the ARP entry 
> manually on host side, so that it doesn't generate an ARP request. May

> be it can help you for a while...
> 
> But for sure this bug needs to be fixed for a use in a real product. I

> woudl appreciate receiving the patch if you find the solution before I

> do.

This can happen if your eCos application uses a different ESA (ethernet
hardware address) than what RedBoot does.  I have seen this - it happens
when the 'fconfig' data isn't set up quite correctly.

> 
> Regards,
> Arnaud.
> 
> -----Message d'origine-----
> De : ecos-discuss-owner@ecos.sourceware.org
> [mailto:ecos-discuss-owner@ecos.sourceware.org] De la part de Jeff 
> Duncan Envoyé : mardi 9 novembre 2004 17:49 À : 
> ecos-discuss@sources.redhat.com Objet : [ECOS] Duplicate ARP table 
> entries
> 
> 
> I'm looking for some insight into a problems I am seeing.  Using the 
> latest eCos with FreeBSD stack on an Arm9 connecting directly to a PC,

> I sometimes get into a situation where communication (such as ping) 
> fails. Looking into it further I have found that there seem to be 
> duplicate entries in the Arp table.  The ethernet output resolves to 
> one entry that has expired triggering an Arp send.  When the Arp 
> response is received, a different Arp table entry is updated with the 
> new Arp timeout.  So what I end up with is a ping request from the 
> host, a response being generated from ICMP, and a new Arp request 
> coming out of the device.
> 
> Sometimes I can fix the problem by clearing the route table and 
> initializing the connection again.
> 
> Any ideas?
> 
> Thanks,
> Jeff
> 
> --
> Before posting, please read the FAQ:
> http://ecos.sourceware.org/fom/ecos
> and search the list archive:
http://ecos.sourceware.org/ml/ecos-discuss
-- 
Gary Thomas <gary@mlbassoc.com>
MLB Associates


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss


--
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss