[lwip-users] Multiple netif interfaces, again.

[Burt]

Adding a second netif interface has resulted in
two mysterious crashes.

The one is in the call to do_delconn where
a connection in a listening state seems to
almost always corrupt one of ->p  (pbuf lists)
in the unsent, unacked, or ooseq tcp_seg lists.

Sometimes the next fields are very small
addresses (as with 0x06) and sometimes they
are larger than the allowed address space.

This crash almost always (but not always)
happens when the socket is closed.

Debug code that periodically walks the
tcp_listen_pcbs list looking for corruption
has failed to nail down the corruption.
That is, the list may be fine at one point
in the code and a few lines later may be
corrupted. Are these lists modified by
multiple threads?

More generally, are there known issues with
using multiple netif interfaces in lwip that
might match what I am experiencing?

Would anyone care to venture forth
possible root causes of these issues?

(Unfortunately, I am forced to use the
lwip distribution from mid-2003 and
cannot simply  update it to the current
one.)

Thank you.
Burt