[PATCH 1/7] [RFC] Support named cgroups hierarchies

Li Zefan lizf at cn.fujitsu.com
Tue Mar 31 23:24:36 PDT 2009

>>> -     /* Next check flags */
>>> -     if (new->flags != root->flags)
>> Is this change intended or unintended? With this change we allow:
>>  # mount -t cgroup -o cpu xxx /mnt1
>>  # mount -t cgroup -o cpu,noprefix xxx /mnt2
>> But files in /mnt2 still prefix with 'cpu.'
> That fits in better with existing semantics for other filesystems - if
> you mount an already-mounted device with different superblock mount
> options, the new mount still has the superblock options associated
> with the old mount.
> Arguably this might mean that we should ignore all options when
> mounting a hierarchy by name that already exists, even if you specify
> a different set of subsystems from what is mounted.

Then I think it's better to make it a different patch, but not hiding
this change in this patch.

>>> +     /* If we asked for subsystems then they must match */
>>> +     if (new->subsys_bits && new->subsys_bits != root->subsys_bits)
>>>               return 0;
>> This has already been checked.
> Where?

My mistake, please ignore. :)

More information about the Containers mailing list