[please wrap your email at about 76 characters; makes it easier to
respond to]
On Fri, 2004-08-13 at 10:09, sandr8 wrote:
> jamal wrote:
> >1) You cant change the enqueue/dequeue API just because one qdisc
> >doesnt seem to use it right. Whats all that contrack stuff doing in
> >dev.c and pkt sched areas?
> >
> the conntrack stuff (actually that little part of code that adds
> unbilling to ACCT in case of a drop) is something that comes later
> (patch 4) and is not in the pkt sched part. please, don't let it
> divert your mind from the real point. just forget it for the moment...
> [in any case it is _not_ in the pkt sched area and as i said in mail
> 4/4 i don't like to put the variables into dev.c, that's why i am
> there asking for alternatives]
This actually seems to be the core issue.
Correct me if i misunderstood what you are trying to achieve:
Somewhere above, the netfilter code bills some packet. Packet gets
all the way to the scheduler on egress.
Scheduler drops packet although it has been billed already.
You being a man looking for justice ;-> decides that was unfair and
you are trying to undo it. Is this accurate?
Also is their a corrective factor that happens once the _accounting_
data has been shipped? Example:
- account for packet
- ship accounting data to some billing server
- oops, unbill
- what now?
BTW, what happens if you clone the packet below netfilter and send
several copies of it possibly over several different interfaces? This
may happen with tc extensions.
I think accounting is important - especially if it is almost free with
contracking.
Lets talk about this issue first instead of confusing it with everything
else you have in other patches.
cheers,
jamal
|