Running two network cards

Hi:

I have been trying to get two network cards running in the same box and
using qnet.

Ultimately I want to be able to see qnet using either card to get to a
destination.

I have two ne2000 cards and am using the big tcpip stack. The cards are
running correctly and I can ping the destination. Also “ls /net/dest”
works.

But : It is using only one of the cards and when I pull out the cable.
All I get is “no route to host”.
This suggests that the second card or the route is bad. To test this I
restart io-net with only a single cable in the second card - all works
ok. In fact all three combinations of 1&2, 1 or 2 work after a restart
but never after a link failure.

The problem seems to be that io-net is not realizing that a card is
disconnected and re-routing to the alternate. Even putting the cable
back in doesn’t get the service back.

Should it be able to do this automatically ?
Is there a way to manually tell it to ‘re-evaluate’ its routing options
?

My rc.local is :
io-net -dne2000 ioport=0x280,irq=7
-dne2000 ioport=0x220,irq=11
-p tcpip
-p qnet resolve=ndp

sleep 2
ifconfig lo0 up
ifconfig en0 172.24.25.60 netmask 255.255.255.0 up
ifconfig en1 172.24.25.62 netmask 255.255.255.0 up

The destination box has an IP of 172.24.25.61 (same netmask).

The routing table says that to get to 172.24.25/24 go through …60.
When I try “route add -net 172.24.25.0 172.24.25.62”, I get file “File
exists”. Is this likely to be part of the problem?

TIA Andy…

NOTE- New phone,fax numbers

Andy Rhind
Symmetry Innovations Pty Ltd.
PO Box 3204
Weston
A.C.T 2611,
Australia
E-Mail: andy@symmetry.com.au
Phone: 61-2-6248 6611
Fax : 61-2-6248 8383

Andy Rhind <camerons_act@bigpond.com> wrote:

Hi:

I have been trying to get two network cards running in the same box and
using qnet.

Ultimately I want to be able to see qnet using either card to get to a
destination.

I have two ne2000 cards and am using the big tcpip stack. The cards are
running correctly and I can ping the destination. Also “ls /net/dest”
works.

But : It is using only one of the cards and when I pull out the cable.
All I get is “no route to host”.
This suggests that the second card or the route is bad. To test this I
restart io-net with only a single cable in the second card - all works
ok. In fact all three combinations of 1&2, 1 or 2 work after a restart
but never after a link failure.

The problem seems to be that io-net is not realizing that a card is
disconnected and re-routing to the alternate. Even putting the cable
back in doesn’t get the service back.

Should it be able to do this automatically ?
Is there a way to manually tell it to ‘re-evaluate’ its routing options
?

My rc.local is :
io-net -dne2000 ioport=0x280,irq=7
-dne2000 ioport=0x220,irq=11
-p tcpip
-p qnet resolve=ndp

sleep 2
ifconfig lo0 up
ifconfig en0 172.24.25.60 netmask 255.255.255.0 up
ifconfig en1 172.24.25.62 netmask 255.255.255.0 up

The destination box has an IP of 172.24.25.61 (same netmask).

I remember that it will be trouble that if you make the
network like that. If en1 in another network, (so your
destination box also need another NIC on another network)
it will work.

Also, you probably could do sth like:

ls /net/dest~sequential:172.24.25.62

To force to go though “172.24.25.62” interface (just make sure
it can go though)

I will re-check the source, and see if I can do anything
for you orignal setting. I am busy in other project now,
but if you haven’t hear me in week, just send me an email
to remind me :slight_smile:

The routing table says that to get to 172.24.25/24 go through …60.
When I try “route add -net 172.24.25.0 172.24.25.62”, I get file “File
exists”. Is this likely to be part of the problem?

Should not, as QNET now auto-detect that the peer is on
same subnet, and will directly using Ethernet Driver
without let the packet go though TCPIP stack.

This remind me, you could also start qnet as:

-p qnet bind=ip

to force QNET using only TCPIP stack (resolve=ndp is default
anyway), and see if there is any difference.

-xtang@qnx.com