audit.36 kernel

[Chris Wright]

* David Woodhouse (dwmw2 at infradead.org) wrote:
> Still doesn't have error handling for the netlink_unicast() call in
> audit_send_reply(). Every time I look at that I just cannot bring myself
> to add the 'skb_get(); /* netlink frees */' which prefixes the
> netlink_unicast() elsewhere. That just looks _wrong_ to me -- there has
> to be something wrong with the refcounting. 

Yeah, the problem is we add our own 'reliable transmission' to netlink.
So, the refcounting is fine in that it's telling the netlink layer that
it's not the only layer that cares about the skb (well, and it works).
The retry is presently based on an audit_buffer which is lacking for
audit_send_reply.

thanks,
-chris