cgroup sysfs entry

cgroup sysfs entry

[David Shwatrz]

Hello,

I have a question about cgroup implementation:

The /sys/fs/cgroup entry is created in kernel/cgroup.c:

static struct kobject *cgroup_kobj;

cgroup_init(void) {
...
cgroup_kobj = kobject_create_and_add("cgroup", fs_kobj);
  if (!cgroup_kobj) {
  err = -ENOMEM;
  goto out;
  }

}

I wondered about this two points:
1) why is the /sys/fs/cgroup entry created at all?
2) why is this done as a sysfs entry.

The cgroup_kobj is not used anywhere in this module apart from
its creation above (and destroying if we later fail with
register_filesystem() in this same method). I don't see anywhere usage
of syfs API to create files underneath
it. In fact, the entries beneath it are created by mkdir of admin or
systemd code. so what's the point of it being a sysfs entry ? wouldn't
simply mkdir be ok ? and why under /sys/fs?

Since cgroups can be mounted by definition on every path, why this is needed?


Regards,
DavidS

Re: cgroup sysfs entry

[Li Zefan]

于 2013/4/26 3:39, David Shwatrz 写道:
> Hello,
> 
> I have a question about cgroup implementation:
> 
> The /sys/fs/cgroup entry is created in kernel/cgroup.c:
> 
> static struct kobject *cgroup_kobj;
> 
> cgroup_init(void) {
> ...
> cgroup_kobj = kobject_create_and_add("cgroup", fs_kobj);
>   if (!cgroup_kobj) {
>   err = -ENOMEM;
>   goto out;
>   }
> 
> }
> 
> I wondered about this two points:
> 1) why is the /sys/fs/cgroup entry created at all?
> 2) why is this done as a sysfs entry.
> 
> The cgroup_kobj is not used anywhere in this module apart from
> its creation above (and destroying if we later fail with
> register_filesystem() in this same method). I don't see anywhere usage
> of syfs API to create files underneath
> it. In fact, the entries beneath it are created by mkdir of admin or
> systemd code. so what's the point of it being a sysfs entry ? wouldn't
> simply mkdir be ok ? and why under /sys/fs?
> 
> Since cgroups can be mounted by definition on every path, why this is needed?
> 

It's supposed to be the standard place to mount cgroupfs.

commit 676db4af043014e852f67ba0349dae0071bd11f3
Author: Greg KH <gregkh-l3A5Bk7waGM@public.gmane.org>
Date:   Thu Aug 5 13:53:35 2010 -0700

    cgroupfs: create /sys/fs/cgroup to mount cgroupfs on

    We really shouldn't be asking userspace to create new root filesystems.
    So follow along with all of the other in-kernel filesystems, and provide
    a mount point in sysfs.

    For cgroupfs, this should be in /sys/fs/cgroup/  This change provides
    that mount point when the cgroup filesystem is registered in the kernel.

    Acked-by: Paul Menage <menage-hpIqsD4AKlfQT0dZR+AlfA@public.gmane.org>
    Acked-by: Dhaval Giani <dhaval.giani-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
    Cc: Li Zefan <lizf-BthXqXjhjHXQFUHtdCDX3A@public.gmane.org>
    Cc: Lennart Poettering <lennart-mdGvqq1h2p+GdvJs77BJ7Q@public.gmane.org>
    Cc: Kay Sievers <kay.sievers-tD+1rO4QERM@public.gmane.org>
    Signed-off-by: Greg Kroah-Hartman <gregkh-l3A5Bk7waGM@public.gmane.org>