In case some of you are having problems with mad clients, here is a
small patch that should reduce the problem. I can certainly appreciate
the reluctance to incorporate such things into 2.0, but I figured a
patch posted to the mailing list might be a nice compromise between
getting 2.0 *done* and having something available for those fortunate
enough to be blessed by these mad clients.
I'm not sure if there should be additional cleanup on the lease and/or
lease->state before returning, but I do know that the server keeps
running even when a client goes nuts. The worst case I can think of is
a lease and/or memory leak when this event occurs (which is better than
a full /var or a dead server).
For more details on the "mad client" problem, see "Code error in WIDE
dhcp client implementation may disrupt ISC dhcp server operation" from
May 12th.
Basically, some clients can get in a state where they broadcast REBINDs
as soon as they get ACKs.
throttle_2.0b1pl27.diff.gz
