[Devel] Re: [PATCH v2 09/11] memcg: propagate kmem limiting information to children

[Glauber Costa]

On 08/17/2012 02:35 PM, Michal Hocko wrote:
> On Fri 17-08-12 14:07:00, Glauber Costa wrote:
>> On 08/17/2012 01:35 PM, Michal Hocko wrote:
>>>>> Above you said "Once enabled, can't be disabled." and now you can
>>>>>>> disable it? Say you are a leaf group with non accounted parents. This
>>>>>>> will clear the flag and so no further accounting is done. Shouldn't
>>>>>>> unlimited mean that we will never reach the limit? Or am I missing
>>>>>>> something?
>>>>>>>
>>>>>
>>>>> You are missing something, and maybe I should be more clear about that.
>>>>> The static branches can't be disabled (it is only safe to disable them
>>>>> from disarm_static_branches(), when all references are gone). Note that
>>>>> when unlimited, we flip bits, do a transversal, but there is no mention
>>>>> to the static branch.
>>> My little brain still doesn't get this. I wasn't concerned about static
>>> branches. I was worried about memcg_can_account_kmem which will return
>>> false now, doesn't it.
>>>
>>
>> Yes, it will. If I got you right, you are concerned because I said that
>> can't happen. But it will.
>>
>> But I never said that can't happen. I said (ok, I meant) the static
>> branches can't be disabled.
> 
> Ok, then I misunderstood that because the comment was there even before
> static branches were introduced and it made sense to me. This is
> inconsistent with what we do for user accounting because even if we set
> limit to unlimitted we still account. Why should we differ here?
> 
There is another thing as well. Mel was right in his comment: I am
actually abusing this bit (because it is flippable), and it seems the
static branch can be updated more than once...

I'll merge your comments, and fix this.