netatalk with appletalk on ppc

netatalk with appletalk on ppc

[Christopher R. Johnson]

Ok, perhaps this should be posted elsewhere, but I figure someone here
might have tried this:

I'm using netatalk - trying to get the atalkd component up.  I've done
this in the past with older kernels, but this time I'm on 2.4.23rc1.
 I'm using the on-chip ethernet support on the 405, and its working
great with IP based stuff.  I've tried two different versions of
netatalk with identical results.

When atalkd starts up it does aarp probing to see if it can come up with
an Appletalk network address.  The probe is sent out on the wire, but
something wierd happens.  the aarp_rcv code gets invoked right away, and
when I look at the packet its examing, its the exact packet that was
just sent out.  This code sees this as an aarp probe being done by some
other network entity (why?  the mac address is mine) and fails the
probe.  The aarp code then moves on to another address to try, and the
whole thing happens again, forever.  Its definitely going to eth0, and
as I mentioned I see the aarp probe go out on the wire.  Running similar
code on my intel desktop works fine, the probes happen 10 times, .1
seconds apart, no one answers, and the code adopts that Appletalk
address (the expected behavior).  I'm trying to wind my way through the
net_tx_action and net_rx_action code to see why this is happening, but
its slow going.

Suggestions?


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: netatalk with appletalk on ppc

[Christopher R. Johnson]

Update - only happens with half-duplex connections.  Apparently the
"reflected" packets are passed along to the receive driver, even tho its
source address matches the local interface.

Christopher R. Johnson wrote:

>
> Ok, perhaps this should be posted elsewhere, but I figure someone here
> might have tried this:
>
> I'm using netatalk - trying to get the atalkd component up.  I've done
> this in the past with older kernels, but this time I'm on 2.4.23rc1.
> I'm using the on-chip ethernet support on the 405, and its working
> great with IP based stuff.  I've tried two different versions of
> netatalk with identical results.
>
> When atalkd starts up it does aarp probing to see if it can come up with
> an Appletalk network address.  The probe is sent out on the wire, but
> something wierd happens.  the aarp_rcv code gets invoked right away, and
> when I look at the packet its examing, its the exact packet that was
> just sent out.  This code sees this as an aarp probe being done by some
> other network entity (why?  the mac address is mine) and fails the
> probe.  The aarp code then moves on to another address to try, and the
> whole thing happens again, forever.  Its definitely going to eth0, and
> as I mentioned I see the aarp probe go out on the wire.  Running similar
> code on my intel desktop works fine, the probes happen 10 times, .1
> seconds apart, no one answers, and the code adopts that Appletalk
> address (the expected behavior).  I'm trying to wind my way through the
> net_tx_action and net_rx_action code to see why this is happening, but
> its slow going.
>
> Suggestions?
>
>
>


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/