[PATCH 1/9] [RFC] Support named cgroups hierarchies

[Paul Menage]

On Wed, Jul 1, 2009 at 7:28 PM, KAMEZAWA
Hiroyuki<kamezawa.hiroyu at jp.fujitsu.com> wrote:
>> Open issues:
>>
>> - should the specification be via a name= option as in this patch, or
>>   should we simply use the "device name" as passed to the mount()
>>   system call?  Using the device name is more conceptually clean and
>>   consistent with the filesystem API; however, given that the device
>>   name is currently ignored by cgroups, this would lead to a
>>   user-visible behaviour change.
>>
>
> IMHO, name= option is good because people think device name for pseudo file
> system has no meanings. I think just leaving it as "no meaning" is better.
>

Yes, I guess that makes sense. That was Li Zefan's opinion too.

>>
>> +#define MAX_CGROUP_ROOT_NAMELEN 64
>> +
>
> I don't like long names but....isn't this too short ? How about NAME_MAX ?
>


>
>>  /*
>>   * A cgroupfs_root represents the root of a cgroup hierarchy,
>>   * and may be associated with a superblock to form an active
>> @@ -93,6 +95,9 @@ struct cgroupfs_root {
>>
>>       /* The path to use for release notifications. */
>>       char release_agent_path[PATH_MAX];
>> +
>> +     /* The name for this hierarchy - may be empty */
>> +     char name[MAX_CGROUP_ROOT_NAMELEN];
>>  };
>>
> If you don't want to make cgroupfs_root bigger,
>
>        cgroupfs_root {
>                ......
>                /* this must be the bottom of struct */
>                char name[0];
>        }
>
> Is a choice.

I'd rather avoid something like that since I think it's less readable
- I'd probably just make the name into a pointer in that case.

>
> BTW, reading a patch, any kind of charactors are allowed ?

Yes, other than \000 of course. I guess maybe I should use
seq_escape() to print the name to avoid confusion in the event that
people put whitespace in there, or else just ban whitespace (or maybe
all non-alphanumeric chars).

Paul