pppoe "taking interface down"

[Paul Ripke <stix%stix.id.au@localhost>]

Since upgrading to netbsd-7 branch kernel, I've had my pppoe interface
taken down by the kernel on a few occasions, and up'ing the interface
is required a few times to actually get it up and running again. It
looks as though maybe the remote end queues a PADT, which I get as
soon as I attempt to reconnect. Previously, my interface has
reconnected automatically just fine, requiring no manual intervention.
I'm hesitant to blame netbsd, since I can't find any relevant changes,
so maybe it's my ISP at fault. However, I'd like that "no intervention,
auto reconnect" behaviour back, if possible.

I did try setting IFF_AUTO (aka IFF_LINK1) but that did not help, I
still got disconnected, and seemed to require deleting and re-creating
the interface to get back to a working state.

I haven't caught disconnect event with debug enabled, however, I have
enabled debug and attempted to reconnect, resulting in:

Oct 28 17:54:10 slave /netbsd: pppoe0: lcp close(initial)
Oct 28 17:54:10 slave /netbsd: pppoe0: lcp open(initial)
Oct 28 17:54:10 slave /netbsd: pppoe0: phase establish
Oct 28 17:54:11 slave /netbsd: pppoe0: connected to lns20.syd4
Oct 28 17:54:11 slave /netbsd: pppoe0: connected to lns20.syd4
Oct 28 17:54:11 slave /netbsd: pppoe0: received unexpected PADO
Oct 28 17:54:11 slave /netbsd: pppoe0: connected to lns20.syd7
Oct 28 17:54:11 slave /netbsd: pppoe0: received unexpected PADO
Oct 28 17:54:11 slave /netbsd: pppoe0: connected to lns20.syd7
Oct 28 17:54:11 slave /netbsd: pppoe0: received unexpected PADO
Oct 28 17:54:11 slave /netbsd: pppoe0: session 0x62f9 connected
Oct 28 17:54:11 slave /netbsd: pppoe0: lcp up(starting)
Oct 28 17:54:11 slave /netbsd: pppoe0: lcp output <conf-req id=0x2a len=14 ...>
Oct 28 17:54:11 slave /netbsd: pppoe0: lcp input(req-sent): <conf-req id=0x1 ...>
Oct 28 17:54:11 slave /netbsd: pppoe0: lcp parse opts: mru auth-proto magic
Oct 28 17:54:11 slave /netbsd: pppoe0: lcp parse opt values:  mru 1492 auth-proto [mine pap != his chap] magic 0x89f44c52 send conf-nak
Oct 28 17:54:11 slave /netbsd: pppoe0: lcp output <conf-nak id=0x1 len=9 ...>
Oct 28 17:54:11 slave /netbsd: pppoe0: lcp input(req-sent): <conf-ack id=0x2a len=14 ...>
Oct 28 17:54:11 slave /netbsd: pppoe0: lcp input(ack-rcvd): <conf-req id=0x2 len=18 ...>
Oct 28 17:54:11 slave /netbsd: pppoe0: lcp parse opts: mru auth-proto magic
Oct 28 17:54:11 slave /netbsd: pppoe0: lcp parse opt values:  mru 1492 auth-proto magic 0x89f44c52 send conf-ack
Oct 28 17:54:11 slave /netbsd: pppoe0: lcp output <conf-ack id=0x2 len=18 ...>
Oct 28 17:54:11 slave /netbsd: pppoe0: lcp tlu
Oct 28 17:54:11 slave /netbsd: pppoe0: phase authenticate
Oct 28 17:54:11 slave /netbsd: pppoe0: pap output <req id=0x8 len=47 ...>
Oct 28 17:54:12 slave /netbsd: pppoe0: pap peer TO
Oct 28 17:54:12 slave /netbsd: pppoe0: pap output <req id=0x9 len=47 ...>
Oct 28 17:54:21 slave /netbsd: pppoe0: lcp input(opened): <term-req id=0x3 len=4>
Oct 28 17:54:21 slave /netbsd: pppoe0: phase terminate
Oct 28 17:54:21 slave /netbsd: pppoe0: lcp send terminate-ack
Oct 28 17:54:21 slave /netbsd: pppoe0: lcp output <term-ack id=0x3 len=4>
Oct 28 17:54:21 slave /netbsd: pppoe0: session 0x62f9 terminated, received PADT
Oct 28 17:54:21 slave /netbsd: pppoe0: lcp down(stopping)
Oct 28 17:54:21 slave /netbsd: pppoe0: Down event (carrier loss), taking interface down.
Oct 28 17:54:21 slave /netbsd: pppoe0: lcp close(starting)
Oct 28 17:54:21 slave /netbsd: pppoe0: phase dead

Anyone seen anything similar? Ideas? Suggestions?

-- 
Paul Ripke
"Great minds discuss ideas, average minds discuss events, small minds
 discuss people."
-- Disputed: Often attributed to Eleanor Roosevelt. 1948.