[PATCH] resolve byte order mess in ip_input/ip_output/pfil(9)

[Ermal Luçi]

On Fri, Oct 5, 2012 at 3:12 PM, Gleb Smirnoff <glebius at freebsd.org> wrote:
>   Ermal,
>
> On Fri, Oct 05, 2012 at 03:01:38PM +0200, Ermal Lu?i wrote:
> E> it would be better to switch to net byte order allover rather than
> E> trade one for the other.
> E> This makes it even more tricky to understand the code than it is.
> E> If you do the work its better to do the full thing in one shot and
> E> switch to netbyte order.
>
> Please read carefully my description and patch. It creates a definite
> points in stack where byte order is swapped. One point where it is
> swapped into host, and one point where it is swapped back into net.
>
> Patch already narrows down the scope of host byte order in the stack,
> host byte order is now between to definite points. If anyone ever wants
> to switch entire stack to net byte order, let it be. Current patch is
> just step in this direction.
>
> The fast forwarding path is already entirely in net byte order. Even
> if run with ipfw and/or pf.
>
> E> speaking of pf(4) side of things please do not loose the VIMAGE calls!
>
> Yeah, can you explain please why do we need them here? The pfil hooks
> are always run already in some defined VNET context, don't they?
>

from my testing at the time these were needed otherwise you will get issues.
I do not remember the details but i put those there because were required.
There is no overhead as well from leaving those there.

> --
> Totus tuus, Glebius.



-- 
Ermal