[Cake] cake review comments cake_drop

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

[-- Attachment #1: Type: text/plain, Size: 1089 bytes --]

Hello All,

On 18th Aug (before I'd subscribed to the list), Dave posted a message
about reviewing cake
https://lists.bufferbloat.net/pipermail/cake/2015-August/000364.html

In it he mentioned a change to cake_drop:

"

I would like cake_drop to be evaluated with some less exaustive
search. It also looks wrong

      for(j=0; j < CAKE_MAX_CLASSES; j++) {
                fqcd = &q->classes[j];

CAKE_MAX_CLASSES should actually be q->class_cnt here; I think. I am
unsure if other references to it are correct in face of changes, on
cleanup"


After making the change and nearly submitting a pull request I've
thought about this some more and to be blunt it makes me nervous.  My
concern is what happens if cake is changed from say a diffserv8 config
to a single class(bin) config and for whatever reason 'cake_drop' is
called.  with diffserv8 the fat flow is likely to be in a class(bin)
higher than the first....cake_drop would now only check the first class
for the fat flow and possibly not find anything to drop....  I fear what
happens next!

Kevin


[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/pkcs7-signature, Size: 4816 bytes --]