Re: [PKT_SCHED]: Allow using nfmark as key in U32 classifier.

[Thomas Graf <tgraf@suug.ch>]

* jamal <1104767726.1049.591.camel@jzny.localdomain> 2005-01-03 10:55
> Sorry, havent looked at the code yet. Is the patch you posted the same
> as you have on the url earlier?

Yes, with a few minor fixes and I made it depend on
CONFIG_NET_CLS_EMATCH and added nil macros when it's not defined.
Those macros need some more work though.

TODO:
 - KConfig doc
 - make change optional
 - dump m->data if no dump implementation is provided

The question I'm yet unsure is whether to strictly return 1/0 or
allow classifier return codes.

Given I'm done with the above, a minimal ematch will look  like:

static int sec_match(struct sk_buff *skb, struct tcf_ematch *m)
{
	return *(u32 *) m->data == skb->security;
}

static struct tcf_ematch_ops sec_ops = {
	.kind = 11,
	.match = sec_match,
	.owner = THIS_MODULE
}

... init/exit calling tcf_em_(un_register ...

What do you think about checking RTA_PAYLOAD of the ematch data
for < sizeof(unsigned long) and assign it to m->data without
allocation? Would save us nonse allocations. The yet unused 8bit
in tcf_ematch_hdr would hold a flag to indicate that m->data
contains the data directly.

> I relaized that - but should the user know anything about this?

Actually no but initializing it in tcf_em_register wouldn't serve
any purpose so we can either remove it or leave it there.