Okay, I looked at the code again and see that it doesn't send the
keepalive if there is activity on the line like you suggest.  I missed
that the first glance through.  However, it also looks like the activity
counter (pp_last_activity) the keepalive mechanism consults is reset in
sppp_output() when the local pppoe0 interface _sends_ a packet.  I don't
understand most of the pppoe stuff, but this seems fishy to me.
Shouldn't it only be reset when the pppoe0 interface receives a packet?

//Jesse Off

On Sun, 24 Aug 2003, Martin Husemann wrote:

> On Sun, Aug 24, 2003 at 10:47:48AM -0700, Jesse Off wrote:
> > I don't have a problem with establishing or reestablishing the session,
> > its just knowing when a session its already got is dead.
>
> Oh, sorry, I did not read carefull enough.
>
> You are right, the change won't help. Is there *any* traffic on the pppoe
> connection? Real traffic will prevent keep-alive LCP echo requests (if IP
> gets through, why should we bother probing LCP?)
>
> Martin
>