[RFC][PATCH 0/7 v2] memcg: prevent failure in pre_destroy()

[KAMEZAWA Hiroyuki <kamezawa.hiroyu-+CUm20s59erQFUHtdCDX3A@public.gmane.org>]

This is a v2 patch for preventing failure in memcg->pre_destroy().
With this patch, ->pre_destroy() will never return error code and
users will not see warning at rmdir(). And this work will simplify
memcg->pre_destroy(), largely.

This patch is based on linux-next + hugetlb memory control patches.

I post this as RFC because I'll have vacation in the next week and
hugetlb patches are not visible in linux-next yet.
So, I'm not in hurry. Please review when you have time.

I'll rebase this onto memcg-devel in the next post.
== BTW, memory cgroup's github is here == 
git://github.com/mstsxfx/memcg-devel.git

Since v1, Whole patch designs are changed. In this version, I didn't
remove ->pre_destroy() but make it succeed always. There are no
asynchronous operation and no big patches. But this introduces
2 changes to cgroup core.

After this series, if use_hierarchy==0, all resources will be moved
to root cgroup at rmdir() or force_empty().

Brief patch conents are

0001 : my version of compile-fix for linux-next, Aneesh will post his own version.
0002 : fix error code in hugetlb_force_memcg_empty
0003 : add res_counter_uncharge_until()
0004 : use res_counter_uncharge_until() at move_parent()
0005 : move charges to root cgroup at rmdir, if use_hierarchy=0
0006 : clean up mem_cgroup_move_account()
0007 : cgroup : avoid attaching task to cgroup where ->pre_destroy() is running.
0008 : cgroup : avoid creating a new cgroup under a cgroup where ->pre_destroy() is running.
0009 : remove -EINTR from memcg->pre_destroy().

Thanks,
-Kame