Re: [Cake] de-natting & host fairness

[Kevin Darbyshire-Bryant <kevin@darbyshire-bryant.me.uk>]

On 28/09/16 04:06, Jonathan Morton wrote:
>
>> On 28 Sep, 2016, at 05:56, Kevin Darbyshire-Bryant
>> <kevin@darbyshire-bryant.me.uk> wrote:
>>
>> Does this need to be another variable/parameter or could it be the
>> next bit along in the flow type?
>
> I’ve already pushed it to the ‘cobalt’ branch, so you can see how
> I’ve done it and start testing.  I’ve verified that it compiles, no
> more than that so far.
>
> For configuration, there is a separate flag parameter passed.
> Internally, I’ve used another bit of the existing flow_mode field
> (but not the next one along).  The latter is also how the
> configuration is read back out again to tc.
>
> Overall, the patch ended up much smaller than the original.  Switch
> statements in C are actually quite verbose.

Looks good and as you say much smaller without the switch stuff and 
IPv6.  As a further bikeshed.....

Is it worth doing:

if (reverse) {
    all the src/dst swaps
    if ports do port src/dst swaps
    nf_ct_put
} else {
    all the dst/src swaps
    if ports do port dst/src swaps
}

Code gets duplicated...or possibly not depending on the compiler, but 
those ternaries are if/else in disguise...and we do a few of them, so if 
we did one at the cost of some duplicate code......



>
> - Jonathan Morton
>