Hi Tejun,

On 17.11.2012 04:30, Tejun Heo wrote:

Hello, guys.

This patchset implements proper hierarchy support for netcls_cgroup.
Pretty simliar to the netprio one[3].  Simpler as each cgroup has
single config value instead of array of them.

This patchset contains the following three patches.

  0001-netcls_cgroup-introduce-netcls_mutex.patch
  0002-netcls_cgroup-introduce-cgroup_cls_state-is_local.patch
  0003-netcls_cgroup-implement-proper-hierarchy-support.patch

This patchset is on top of

cgroup/for-3.8 ef9fe980c6 ("cgroup_freezer: implement proper hierarchy support")
+ [1] "[PATCHSET cgroup/for-3.8] cgroup: allow ->post_create() to fail"
+ [2] "[PATCH 1/2] cgroup: s/CGRP_CLONE_CHILDREN/CGRP_CPUSET_CLONE_CHILDREN/"
       "[PATCH 2/2] cgroup, cpuset: remove cgroup_subsys->post_clone()"
+ [3] "[PATCHSET cgroup/for-3.8] netprio_cgroup: implement hierarchy support"

and available in the following git branch.

  git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup.git review-netcls_cgroup-hierarchy

diffstat follows.

  include/net/cls_cgroup.h |    1
  net/sched/cls_cgroup.c   |  102 ++++++++++++++++++++++++++++++++++++++++-------
  2 files changed, 88 insertions(+), 15 deletions(-)

Thanks.

I played a bit around with this series, e.g. creating a hierarchy and and changing 
the root classid and seeing the change propagating through the hierarchy. Everything
worked as described in the documentation.

Sorry to bring this up again: how should to root cgroup behave? If the ultimate
goal to have only one single hierarchy then I would assume it is important that
the semantic for all controllers are the same. As you pointed out the networking
controllers are kind of a strange beast in the zoo of the cgroup controllers. 
But still I would assume that all root controllers behave the same. memcg or cpu*
are not expected to do any work in the root cgroup.

cheers,
daniel