Re: [PATCH] tcp: assign the sock correctly to an outgoing SYNACK packet

[Paul Moore <pmoore@redhat.com>]

On Monday, April 08, 2013 05:33:25 PM David Miller wrote:
> From: Paul Moore <pmoore@redhat.com>
> Date: Mon, 08 Apr 2013 17:24:50 -0400
> 
> > If the void pointer is wrapped by a #ifdef (plenty of precedence for that)
> > and the management of that pointer is handled by LSM hooks why is it a
> > concern?  I apologize for pushing on the issue, but I'm having a hard
> > time reconciling the reason for the "no" with the comments/decisions
> > about the regression fix; at present there seems to be a level of
> > contradiction between the two.
>
> 8 bytes times however many millions of packets per second we can process
> on a big machine, you do the math.
> 
> It's memory, less cache locality, etc. etc. etc.
> 
> It's the most important data structure in the entire networking stack,
> and every single byte matters.
> 
> I want the overhead to be your problem, so that only users of your
> stuff eat the overhead, rather than everyone.

Okay, if the objection is really just one of structure size and not the hooks, 
what if I did the work to consolidate the skb->secmark and skb->sp fields into 
a new structure/pointer?  Assuming it wasn't too painful, it would be a net 
reduction of four bytes.  If that worked would you have an objection to us 
adding a LSM security blob to this new structure?

-- 
paul moore
security and virtualization @ redhat