Re: [RFC][PATCH 5/4] memcg: don't call res_counter_uncharge when obsolete

[Balbir Singh <balbir@linux.vnet.ibm.com>]

* KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> [2009-01-15 13:41:14]:

> On Thu, 15 Jan 2009 09:47:50 +0530
> Balbir Singh <balbir@linux.vnet.ibm.com> wrote:
> 
> > * KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> [2009-01-15 12:24:16]:
> 
> > > But I don't like -EBUSY ;)
> > > 
> > > When rmdir() returns -EBUSY even if there are no (visible) children and tasks,
> > > our customer will take kdump and send it to me "please explain this kernel bug"
> > > 
> > > I'm sure it will happen ;)
> > >
> > 
> > OK, but memory.stat can show why the group is busy and with
> > move_to_parent() such issues should not occur right? I'll relook at
> > the code. Thanks for your input.
> > 
> 
> There was a design choice at swap_cgroup.
> 
> At rmdir, there may be used swap entry in memcg. (mem->memsw.usage can be > 0)
>   1. update all records in swap cgroup
>   2. just ignore account from swap, we can treat then at swap-in.
> 
> I implemented "2" by refcnt.
> 
> To do "1", we have to scan all used swap_cgroup but I don't want to scan all
> swap_cgroup entry at rmdir. It's heavy job.
> (*) To reduce memory usage by swap_cgroup, swap_cgroup just have a pointer to memcg
> (**) I implemented swap_cgroup as statically allocated array because I don't want
>     any dynamic memory allocation at swap-out and want to avoid unnecessary memory
>     usage.

Fair enough, but I don't like that we don't have any checks for

If parent still has children, parent should not go away. The
problem that Daisuke-San is seeing.

-- 
	Balbir