netdev - Re: [RFC PATCH 0/5] IPsec parallelization

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20081202075804.GD13998@secunet.com>
Date:	Tue, 2 Dec 2008 08:58:04 +0100
From:	Steffen Klassert <steffen.klassert@...unet.com>
To:	Herbert Xu <herbert@...dor.apana.org.au>
Cc:	David Miller <davem@...emloft.net>, netdev@...r.kernel.org,
	klassert@...hematik.tu-chemnitz.de
Subject: Re: [RFC PATCH 0/5] IPsec parallelization

On Mon, Dec 01, 2008 at 07:15:01PM +0800, Herbert Xu wrote:
> On Mon, Dec 01, 2008 at 02:29:29AM -0800, David Miller wrote:
> >
> > Herbert, I'm not completely convinced of this line of thinking :-)
> 
> Well you've put some doubts into mind for a second there, but
> after thinking some more I'm sticking to my position :)
> 
> > Will crypto be faster than a routing cache lookup?  Because flow
> > seperation helps routing performance, significantly.
> 
> No to the first part, but that's not the point.  I'm saying that
> you can do flow separation within the crypto layer, well, for
> output anyway.  All you need to do is to provide some key to
> the crypto layer that tells it that ordering is only required
> within that key.  Alternatively, just allocate plenty of crypto
> tfms as these patches do, and feed the flows down to different
> tfms.
> 
> For input there's nothing we can do apart from processing completed
> requests in sequence so no flow information is going to help (and
> in fact if it did one could argue that the purpose of IPsec has been
> defeated :)
> 
> > And the reconstitution of the seperated crypto operations into the
> > original properly ordered flow can be done most cleanly (IMHO) in the
> > networking layer as far as I have seen so far.
> 
> Well once I've finished the GRO stuff I can give this a go.  But
> I'm pretty sure we can achieve exactly the same thing with only
> minimal changes to xfrm.  Really all it needs to do is allocate
> more tfm objects for the output path.
> 
> All the serialisation logic can stay in the crypto layer because
> AFAICS no network-specific knowledge is being used in this patch
> set.
> 

The parallelization/serialization interface is generic, so the crypto
layer can use it in the same way as xfrm.

The only network specific information that is used is the knowledge
of the xfrm_state that a packet uses. On serialization, this helps
to stay as parallel as possible. Packets that uses the same state must
go to the same cpu (because of the replay window) packets that use
different states can go to different cpus.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html