Re: [PATCH nf-next v3] netfilter: xtables: lightweight process control group matching

[Daniel Borkmann <dborkman-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>]

On 12/28/2013 10:20 AM, Li Zefan wrote:
> On 2013/12/28 16:44, Daniel Borkmann wrote:
>> On 12/28/2013 09:17 AM, Daniel Borkmann wrote:
>>> On 12/28/2013 02:33 AM, Li Zefan wrote:
>>>> On 2013/12/27 16:05, Daniel Borkmann wrote:
>>>>> On 12/27/2013 04:13 AM, Li Zefan wrote:
>>> ...
>>>>>> I'd suggest splitting cls_cgroup code into 2 parts. The first part
>>>>>> is to manage cgroupfs and classid, and should be put into net/core/
>>>>>> and add a new config like NET_CGROUP_CLASSID for it. The second part
>>>>>> is specific cls_cgroup code.
>>>>>
>>>>> Sure, if this is wished, I'd do this as a follow-up as it doesn't affect
>>>>> any of this code in netfilter here.
>>>>
>>>> We should clean up the code before introducing a new feature, not the
>>>> other way.
>>>
>>> Hehe, quite honestly, I think this is YOUR opinion of a "cleanup", which
>>> makes the code actually more complicated, and it's not strictly needed.
>>>
>>> If you are so desperate about this separation, fine, I will do this, but
>>> again, I don't think it's *strictly* required for this.
>>
> 
> Currently we have:
> 
> 	cls_cgroup
> 
> With your patch, it's changed to:
> 
>                                 __  cls_cgroup
>                               /
> 	net_cgroup_classid -
>                               \ __  netfilter_cgroup
> 
> The code should be re-organized.
> 
>> Thinking about this further a bit, with this separation into net/core/, you
>> would need another Kconfig, where people choose between built-in or
>> module (or none).
> 
> Just disable module support for net_cgroup_classid. I never like the cgroup
> feature that allows modular subsystems. It will be just about 100 lines of
> code, so we gain little by making it modular. Futhermore with this change,
> we can simplify include/linux/cls_cgroup.h
> 
> Moreoever cls_cgroup is the only modular cgroup subsystem, so I may even remove
> this feature from cgroup core after the change to cls_cgroup.

Ok, will send a v4 with this change included, thanks.

>> With built-in you would disallow people to load/unload the
>> cgroup part during runtime. With the module, you would introduce a module
>> dependency _for each_, even if you only want to use net_cls, which sort of
>> is not what we want from a simple "cleanup".