[Cake] Cake Diffserv8 Incorrect?

[Cake] Cake Diffserv8 Incorrect?

[Noah Causin]

Hi,

I've started using diffserv8 mode on cake with packet marking, and I 
have been noticing that traffic does not appear to go into its correct 
tin in tc -s qdisc in LEDE.

This is what I notice:

cs0 in tin 2
cs1 in tin 0
cs2 in tin 2
cs3 in tin 3
cs4 in tin 6
cs5 in tin 6
cs6 in tin 7
cs7 in tin 7

Noah Causin

Re: [Cake] Cake Diffserv8 Incorrect?

[Jonathan Morton]

[-- Attachment #1: Type: text/plain, Size: 178 bytes --]

That looks normal to me. If you really want a 1:1 mapping from CS to tin,
you should use "precedence". But that's an obsolete interpretation of the
TOS field.

- Jonathan Morton

[-- Attachment #2: Type: text/html, Size: 231 bytes --]

Re: [Cake] Cake Diffserv8 Incorrect?

[Noah Causin]

Thank you.

I just assumed that diffserv8 matched up with the CS classes.

On 11/6/2016 1:04 AM, Jonathan Morton wrote:
>
> That looks normal to me. If you really want a 1:1 mapping from CS to
> tin, you should use "precedence". But that's an obsolete
> interpretation of the TOS field.
>
> - Jonathan Morton
>

Re: [Cake] Cake Diffserv8 Incorrect?

[Sebastian Moeller]

Hi Noah,

So far the only way to figure out what cake is supposed to do is looking at the source code, as there are no high-level design documents, nor even a proper man page. That said I believe that cake's diffuser modes correspond to Jonathan's interpretation of different related RFCs; not that these RFCs necessarily make much sense and it is not clear which RFC a specific mode was based on. Since DSCPs are only ever valid inside a DSCP-domain like a home net, I guess most of the schemes are workable as long as the mapping is well documented. Now it would be sweet if cake would grow a robust and comprehensive documentation, and while I believe this is going to happen, I would not hold my breath...

Best Regards
       Sebastian

On November 6, 2016 5:49:56 AM GMT+01:00, Noah Causin <n0manletter@gmail.com> wrote:
>Hi,
>
>I've started using diffserv8 mode on cake with packet marking, and I 
>have been noticing that traffic does not appear to go into its correct 
>tin in tc -s qdisc in LEDE.
>
>This is what I notice:
>
>cs0 in tin 2
>cs1 in tin 0
>cs2 in tin 2
>cs3 in tin 3
>cs4 in tin 6
>cs5 in tin 6
>cs6 in tin 7
>cs7 in tin 7
>
>Noah Causin
>_______________________________________________
>Cake mailing list
>Cake@lists.bufferbloat.net
>https://lists.bufferbloat.net/listinfo/cake