Re: [RFC PATCH 0/2] Faster/parallel SYN handling to mitigate SYN floods

[Jesper Dangaard Brouer <jbrouer@redhat.com>]

On Thu, 2012-05-31 at 14:58 +0200, Eric Dumazet wrote:
> On Thu, 2012-05-31 at 14:51 +0200, Jesper Dangaard Brouer wrote:
> > On Thu, 2012-05-31 at 00:40 +0200, Jesper Dangaard Brouer wrote:
> > > That seems like a very unlikely situation, which we perhaps should
> > > neglect as we are under SYN attack.
> > >
> > > I will test the attack vector, if we instead of dropping the reqsk,
> > > fall back into the slow locked path.
> > 
> > I can provoke this attack vector, and performance is worse, if not
> > dropping the reqsk early.
> > 
> > Generator SYN flood at 750Kpps, sending false retransmits mixture.
> > 
> > - With early drop: 406 Kpps
> > - With return to locked processing: 251 Kpps
> > 
> > Its still better than the approx 150Kpps, without any patches.
> > 
> 
> How many different IP addresses are used by your generator ?

In this attack I reduced the IPs to 255, and also the source port
numbers, and then simply cloned some of the SKBs.  But normally I use
65535 IPs 198.18.0.0/16 (the range reserved for benchmarking).


> Or maybe you disabled IP route cache ?

Why do you think I have disabled the IP dst route cache?