Why Linux is blind to this ARP reply 2971

Thanks for jumping in!!!

no cross, straight, my mistake.

Firewall issue: no firewall, I made sure of that. Also, this linux box (inside of vmware) can perfectly ping any other host inside my network (from 10.10.0.0 to 10.10.7.254) and replies to pings from all hosts I tried in the same network).

All the pings are using IP address, no host names.

When you ping from linux (10.10.3.99) you get this: # ping 10.10.0.46 PING 10.10.0.46 (10.10.0.46) 56(84) bytes of data. . . and forever goes like that.

route table exactly like you said, plus some extra entries for extra subnets. But the one of interest is there: 10.10.0.0 * 255.255.248.0 U 0 0 eth0

? Yes indeed, I can see (from the windows machine, running ethereal) that the router issued the ARP reply, after receiving the ARP request from the Linux box. It is Linux who is blind to that request. In one test I used a HUB, instead of direct connection, I plug 2 windows machines (10.10.3.5 and 10.10.3.112 )and my Laptop with vmware LInux (10.10.3.99). The 2 windows machine running ethereal can see the Router replying to the ARP reply in response to 10.10.3.99 (linux) ARP request. Tcpdump in the linux does not show a thing coming from 10.10.0.46, nothing, no broadcasts, no unicast to somebody else, nothing,

In the case of the HUB, if I do:

#ping 10.10.3.112 and in another Linux virtual console I have tcpdump host 10.10.3.112 I got this: tcpdump: listening on eth0 datestamp arp who-has 10.10.3.112 tell 10.10.3.99 datestamp arp reply 10.10.3.112 is-at x:x:x:x:x:x

...... If now I do a #ping 10.10.0.46 and in the other console I have: #tcpdump host 10.10.0.46 tcpdump: listening on eth0 datestamp arp who-has 10.10.0.46 tell 10.10.3.99 datestamp arp who-has 10.10.0.46 tell 10.10.3.99 . . and forever like this

Before issuing the previous ping, I launched Ethereal in the Windows machine connected to the same HUB where 10.10.0.46, and 10.10.3.99 are connected to. And Ethereal reports:

No. Time Source Destination Protocol Info 1 0.0000000 10.10.3.99 Broadcast ARP Who has 10.10.0.46? Tell 10.10.3.99 2 0.000442 10.10.0.46 10.10.3.99 ARP 10.10.0.46 is at 02:00:22:10:46:00

Note frame number 2, it is the good router answering Linux its own request.

I have two of these routers (2210), 10.10.0.46 and 10.10.0.39, in both units I have the same behaviour. Actually I am more tempted in saying that: with both routers Linux presents the same behaviour.

IPs of Windows and Linux so far apart and different (just curious -- should make no difference)? Are you sure the router is entered as Linux's default route-GW?

Both, the Windows and the Linux guest (vmware guest) are using DHCP. Of course I jumped to static IP when using the direct connection to the router.

These Routers are not to be used to route any segment that the linux box should be aware of. I installed them with other purpose in mind. We discovered that there was a problem when people connecting to our network from outside (through a VPN channel with Netscreen Firewall) could not see the Routers. I told them, I had tested them (the routers) but further investigation showed that the Firewall could not get a ping reply (echo reply) from the routers. Then I switched to Linux (a server I have running on a dedicated machine, not the vmware), and it behaved the same way than the netscreen firewall (this server is running also 2.4.18.20 from RedHat 9). Then I came to my Laptop and vmware, and you know the rest of the history.

I am sorry for such a long post.

Thanks a lot, Andres