Disabling automatic loopback at kernel level...

Disabling automatic loopback at kernel level...

Post by Venkate » Thu, 08 Apr 2004 02:42:12



Hi,
 I am testing a new Network Interface card, and I need to do
 basic loopback tests on the card. To do this, i have made the
 following configurations:

 - Created two IP aliases on eth1 and eth1:1, and setup two
 IP addresses on different networks [192.50.50.200, and 192.60.60.200]
 on the same interface on my machine.

 - Created two "phantom" IP addresses. Basically, these addresses
 do not exist, but i tried this thinking the kernel may send out
 the packets [192.50.50.10 and 192.60.60.10]

 - Statically added the arp entries of both these phantom ip addresses,
 so that the kernel does not send out an ARP for these addresses.
 [arp -s 192.50.50.10 00:11:22:33:44:55
 arp -s 192.60.60.10 00:11:22:33:44:55]

  Now, when i tried to ping from 192.50.50.200 to 192.60.60.200, tcpdump
 did not capture any packets on eth1.

 Next, i tried to add static routes
 route add -host 192.50.50.200 gw 192.60.60.10
 route add -host 192.60.60.200 gw 192.50.50.10

 [basically trying to make the kernel think that the addresses can be
 reached by the gateways provided.

 Still, tcpdump does not capture packets on eth1.

 Looks like the kernel is recognising that both 192.50.50.200 and
 192.60.60.200 are aliased addresses to the same phy, and it is
 not sending out the packets.

 How do i make the kernel to send packets out to my hardware, so that
 i can loop it back at the hardware?

 Any help will be greatly appreciated!

thanks,
Venkat.

 
 
 

Disabling automatic loopback at kernel level...

Post by Cameron Ker » Thu, 08 Apr 2004 10:50:15



> Hi,
> I am testing a new Network Interface card, and I need to do
> basic loopback tests on the card. To do this, i have made the
> following configurations:

You could try sending broadcast packets out the interface, and trying to
receive them.

Quote:> How do i make the kernel to send packets out to my hardware, so that
> i can loop it back at the hardware?

A better/easier method would be to use two interfaces (or two machines)
and send traffic between them, watching the ifconfig error counters.
(requires a known-good NIC as a test partner however.

Or you could just make yourself a loopback cable (just a RJ45 and two
pairs) and use the diagnostic tools that come with the card.

--
Cameron Kerr

Empowered by Perl!

 
 
 

1. Kernel level 95/NT like registry level functions?

Has anyone done any work on developing kernel level MS registry like
functions?

The benefits would be great in that there would no longer be a need for
software developers to write config file routines but just get the
registry data.

Also this would improve performance because the data IO would be at
kernel level.  There could even be a registry partition like swap to get
rid of ext2 filesystem overhead.

The only problems would be that developers could be hesitant about
developing something that is linux proprietary but since the source
would be PD we could use that to convince SGI or SUN to use it as well.

--

--LINUX- The choice of a GNU generation!--

Kevin Burton
Network Support
General Physics Corporation
http://www.gl.umbc.edu/~kburto1/kburto1.html

"In a weak sense, it [windows NT] is a
form of Unix."  --Bill Gates, 1996

"It is error alone that needs
support of the government."
--Thomas Jefferson

2. c++ and dlopen()

3. Calling kernel level function from user level program

4. Pascal source

5. How can I make the signal solution....to user level..from kernel level

6. libc-4.3 and gcc-2.3.3 on kernel .98p5 ?

7. application level vs kernel level coding ?

8. help: remove java re in redhat

9. questions about link-level loopback, PF_PACKET and ETH_P_LOOP

10. how to disable socket loopback

11. loopback interface set to disabled

12. Automatic level 0 dumps at night?

13. Disabling automatic message retrieval