[ckrm-tech] [PATCH 7/7] containers (V7): Container interface to nsproxy subsystem

Pavel Emelianov xemul at sw.ru
Fri Apr 6 05:59:25 PDT 2007


Paul Menage wrote:
> On 4/5/07, Pavel Emelianov <xemul at sw.ru> wrote:
>>
>> Right now I do not. Calling container_init_eary() is OK from
>> my POV. What I do not like is that I have to reinitialize
>> container->create() callback.
> 
> I don't think you should need to do that. The way cpuset_create()
> handles it is for the root cpuset to be statically allocated, so the
> code looks like:
> 
>     if (!cont->parent) {
>         /* This is early initialization for the top container */
>         set_container_cs(cont, &top_cpuset);
>         top_cpuset.css.container = cont;
>         top_cpuset.mems_generation = cpuset_mems_generation++;
>         return 0;
>     }
> 
>        ... other normal code that allocates a cpuset ...
> 
> Mildly ugly, but it solves the problem.

Well, it's less ugly than resetting pointer but it's worse from
performance POV. See, we have an if (xxx) that is true only once
during system lifetime. This is not that good...

Anyway. When are you going to send your patches with containers
and what kernel will it be built upon? I'll try to prepare the RSS
patches soon after this.

> Paul
> 



More information about the Containers mailing list