On Fri, Aug 20, 2010 at 11:10 PM, Eric Dumazet <eric.dumazet@xxxxxxxxx> wrote:
> Le vendredi 20 août 2010 à 22:53 +0800, Changli Gao a écrit :
>> Since we don't change the tuple in the original direction, we can save it
>> in ct->tuplehash[IP_CT_DIR_REPLY].hnode.pprev for __nf_conntrack_confirm()
>> use.
>>
>> __hash_conntrack() is split into two steps: ____hash_conntrack() is used
>> to get the raw hash, and __hash_bucket() is used to get the bucket id.
>>
>> In SYN-flood case, early_drop() doesn't need to recompute the hash again.
>>
>> Signed-off-by: Changli Gao <xiaosuo@xxxxxxxxx>
>> ---
>
> Hmm... so to accept a few more SYN packets per second in SYNFLOOD
> attack, we slow a bit normal operations ?
SYN-flood case is just a side effect. What I want to do is eliminating
the second call to hash_conntrack() of the original tuple in
__nf_conntrack_confirm().
> (adding one test on each
> packet going through conntrack)
Do you mean the rnd test?
>
> If yes (I dont think we should, hackers are stronger than you anyway,
> just face it)
>
> v4:
> __read_mostly on nf_conntrack_rnd
>
>
> What would happen if we let the initialization of nf_conntrack_rnd
> only in the insertion case (like currently done) ?
> Only the first packet received on the machine/conntrack might be hashed
> on a wrong slot. Is it a big deal ? If yes, maybe find a way to
> recompute the hash in this case, instead of reusing 'wrong' one ?
>
I should keep the old way, but fix a race.
if (unlikely(!nf_conntrack_hash_rnd_initted)) {
get_random_bytes(&nf_conntrack_hash_rnd,
sizeof(nf_conntrack_hash_rnd));
nf_conntrack_hash_rnd_initted = 1;
}
nf_conntrack_alloc() isn't called with in the nf_conntrack_lock. So
the above code maybe executed more than once on different CPUs. It is
easy to fix with the cmpxchg() trick.
Thanks.
--
Regards,
Changli Gao(xiaosuo@xxxxxxxxx)
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
