On 04/11/15 11:09, Toke Høiland-Jørgensen wrote:
> Dave Taht <dave.taht@gmail.com> writes:
>
>> in the general case, I recomend looking at the most current fq_codel
>> code to see how to do it as right as possible. There were several bugs
>> found and fixed in it as well....
> Well, I did, and it doesn't seem to be doing anything different. So much
> so that I tried the same experiment on fq_codel:
>
> $ sudo tc qdisc del dev eno1 root
> $ sudo tc qdisc replace dev eno1 root fq_codel target 100ms interval 200ms
> $ sudo tc qdisc replace dev eno1 root fq_codel target 5ms                 
> $ tc qdisc
> qdisc fq_codel 8007: dev eno1 root refcnt 2 limit 10240p flows 1024 quantum 1514 target 5.0ms interval 200.0ms ecn 
> $ sudo tc qdisc del dev eno1 root                                         
> $ sudo tc qdisc replace dev eno1 root fq_codel target 5ms
> $ tc qdisc                                          
> qdisc fq_codel 8008: dev eno1 root refcnt 2 limit 10240p flows 1024 quantum 1514 target 5.0ms interval 100.0ms ecn 
>
>
> I.e. fq_codel suffers from exactly the same problem.
>
> Is this a bug or is it expected behaviour? I'd say bug?
>
> -Toke
> _______________________________________________
> Cake mailing list
> Cake@lists.bufferbloat.net
> https://lists.bufferbloat.net/listinfo/cake
Doesn't the comment in the code say in essence "if it's the same type of
qdisc then the operation is a change, else the operation is (an atomic)
remove/add" ?  Or did I mis-read it. 

Kevin