Messages in this thread | | | Date | Wed, 12 Aug 2009 22:49:40 +0530 | From | Balbir Singh <> | Subject | Re: [PATCH] Help Resource Counters Scale better (v4.1) |
| |
* KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> [2009-08-13 01:28:57]:
> Balbir Singh wrote: > > Hi, Andrew, > > > > Does this look better, could you please replace the older patch with > > this one. > > > > 1. I did a quick compile test > > 2. Ran scripts/checkpatch.pl > > > > In general, seems reasonable to me as quick hack for root cgroup. > thank you. > > Reviewed-by: KAMEAZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> >
Thanks, yes, we still need to do the percpu counter work, but this will give us breathing space to do it correctly and define strict and non-strict accounting.
> Finally, we'll have to do some rework for light-weight res_counter. > But yes, it will take some amount of time. > My only concern is account leak, but, if some leak, it's current code's > bug, not this patch. >
Yeah.. we'll need to check for that.
> And..hmm...I like following style other than open-coded. > == > int mem_coutner_charge(struct mem_cgroup *mem) > { > if (mem_cgroup_is_root(mem)) > return 0; // always success > return res_counter_charge(....) > } > == > But maybe nitpick. >
Yes, we can refactor for simplication, but we'll need some exceptions. I'll do that as an add-on patch.
-- Balbir
| |