Add ECN support to RED

This code review adds ECN support to RED queue disc. It is based on the changes in code review 306660043 (except for the changes to the RED queue disc therein). This patch should also address bug #2485.

[Tom Henderson, 2016-11-18 04:21:25]

I am fine with this if Shravya, Mohit, and Stefano converge on this solution.  I
don't have an opinion either way about adding UseHardDrop attribute.

[Stefano Avallone, 2016-11-18 08:45:14]

After Mohit mentioned that you discussed what to do (i.e., to drop or to mark)
with packets enqueued when the average queue length exceeds the max threshold
and ECN is enabled, I had another look at the Linux implementation and noticed
the use of the use_hard_drop parameter. So, I believe introducing the
UseHardDrop attribute makes the ns-3 implementation closer to the Linux one and
gives the possibility to decide whether to drop or to mark packets enqueued when
we are above the max threshold. By default, UseHardDrop is false, which means
that such packets are dropped even when ECN is enabled.

[Stefano Avallone, 2016-11-18 08:47:10]

I meant:

By default, UseHardDrop is true, which means that such packets are dropped even
when ECN is enabled.

[Tom Henderson, 2016-11-18 13:56:38]

On 2016/11/18 08:47:10, Stefano Avallone wrote:
> I meant:
> 
> By default, UseHardDrop is true, which means that such packets are dropped
even
> when ECN is enabled.

Yes, that default is what I believe RFC 3168 recommends.  My comment about it
was that I was neutral about exposing the change of such default as an
attribute.

[Mohit Tahiliani, 2016-11-18 16:13:09]

On 2016/11/18 13:56:38, Tom Henderson wrote:
> On 2016/11/18 08:47:10, Stefano Avallone wrote:
> > I meant:
> > 
> > By default, UseHardDrop is true, which means that such packets are dropped
> even
> > when ECN is enabled.
> 
> Yes, that default is what I believe RFC 3168 recommends.  My comment about it
> was that I was neutral about exposing the change of such default as an
> attribute.

Hi Stefano,

These changes look good to me.

Thanks,
Mohit P. Tahiliani

[Shravya, 2016-11-18 17:45:45]

Hello Stefano Sir,
The changes look good to me too. I believe I need to modify the test suite
to check for hard-mark.

Thank you, Sir.


On Fri, Nov 18, 2016 at 9:43 PM, <tahiliani@nitk.edu.in> wrote:

> On 2016/11/18 13:56:38, Tom Henderson wrote:
>
>> On 2016/11/18 08:47:10, Stefano Avallone wrote:
>> > I meant:
>> >
>> > By default, UseHardDrop is true, which means that such packets are
>>
> dropped
>
>> even
>> > when ECN is enabled.
>>
>
> Yes, that default is what I believe RFC 3168 recommends.  My comment
>>
> about it
>
>> was that I was neutral about exposing the change of such default as an
>> attribute.
>>
>
> Hi Stefano,
>
> These changes look good to me.
>
> Thanks,
> Mohit P. Tahiliani
>
> https://codereview.appspot.com/314120044/
>