Re: Multiple NICs - custom protocol development

First, thanks for the response; It's nice to see some community support.

Here is what I am trying to do:

I am building a custom MAC protocol for a wireless system that has different
software on
the "head end" and the "clients." It is not peer-to-peer, While the
hardware is being developed
I want to use Ethernet as a physical layer.

So,I want to use one card running server code and one card running client
code initially. Later I will do
the checkout with multiple client machines and a single server.

If the OS "loops a packet back" (At the IP layer) before it gets to my "MAC
layer" then I can't test any code.

-- Len





On 9/20/07, Lowell Gilbert <freebsd-questions-local@xxxxxxxxxxxxxxx> wrote:

"Len Gross" <sandiegobiker@xxxxxxxxx> writes:

I have a host on my local 192.168.0 / 24 subnet that works fine in
getting
to the Internet via a default route.via a wireless connection.
I want to develop some custom link protocols and I have placed two
Ethernet
NICs in the box.
I want to be able to send packets from one NIC to the other and maintain
the
link to the Internet.
I've tried a large number of things via rc.conf but when I ping of the
cards
it is not going out the interface; it just gets looped back. (I test
this
by disconnecting the "crossover cable" between the two cards.)

My current rc.conf has the following attempt, but this fails.

#
router_enable="Yes"
gateway_enable="Yes"
# Ethernet 1:
ifconfig_xl0="inet 192.168.1.1 netmask 255.255.255.0"
# Ethernet 2
ifconfig_rl0="inet 192.168.2.1 netmask 255.255.255.0"
#
# Set up loop between the two ethernet cards
static_routes "xtor, rtox"
route_rtox = "-host 192.168.1.1 192.168.2.1"
route_xtor = "-host 192.168.2.1 192.168.1.1"

Can I do what I want or must I have a second development box?

What you want to do doesn't make sense; there is no reason to send
packets to yourself over a wire. If your machine is sending packets
to itself, the best path is over the loopback, and it doesn't make
sense to send it over a different path. So you need to examine *why*
you want to do that before you can figure out the best approach to
your root problem.

I do protocol development and testing through a number of different
approaches, but for basic development there's usually no problem with
letting the packets go over the loopback. For working on something
like DHCP, I need separate IP stacks, because that will modify the
routing tables differently on the server and the client(s). For that,
I find virtual machines (qemu, most recently) to be the easiest and
most flexible environment. I have also used environments based on
bpf(4) interfaces when I was working with IP stacks that ran
separately from the system's kernel.

Good luck.

_______________________________________________
freebsd-questions@xxxxxxxxxxx mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-questions
To unsubscribe, send any mail to "freebsd-questions-unsubscribe@xxxxxxxxxxx"

Relevant Pages

  • Re: Sniffer Cant see cluster traffic
    ... None of the original Digital designed Ethernet protocols (SCS, LAT, ... protocol specific multicasts. ... IP over Ethernet on the other hand is full of Ethernet broadcasts. ... I got to see many SCS packets that were specifically adressed ...
    (comp.os.vms)
  • TCP/IP Applications FAQ
    ... waiting for an ICMP Echo Reply from the host. ... Each trio of packets 'expire' at a succeeding ... Of the rexec protocol. ...
    (comp.unix.questions)
  • TCP/IP Applications FAQ
    ... waiting for an ICMP Echo Reply from the host. ... Each trio of packets 'expire' at a succeeding ... Of the rexec protocol. ...
    (comp.unix.questions)
  • TCP/IP Applications FAQ
    ... waiting for an ICMP Echo Reply from the host. ... Each trio of packets 'expire' at a succeeding ... Of the rexec protocol. ...
    (comp.unix.questions)
  • TCP/IP Applications FAQ
    ... waiting for an ICMP Echo Reply from the host. ... Each trio of packets 'expire' at a succeeding ... Of the rexec protocol. ...
    (comp.unix.questions)
Loading