[Devel] Re: [PATCH 16/23] slab: provide kmalloc_no_account

[KAMEZAWA Hiroyuki]

(2012/04/25 23:29), Glauber Costa wrote:

> On 04/24/2012 10:44 PM, KAMEZAWA Hiroyuki wrote:
>> (2012/04/23 8:53), Glauber Costa wrote:
>>
>>> Some allocations need to be accounted to the root memcg regardless
>>> of their context. One trivial example, is the allocations we do
>>> during the memcg slab cache creation themselves. Strictly speaking,
>>> they could go to the parent, but it is way easier to bill them to
>>> the root cgroup.
>>>
>>> Only generic kmalloc allocations are allowed to be bypassed.
>>>
>>> The function is not exported, because drivers code should always
>>> be accounted.
>>>
>>> This code is mosly written by Suleiman Souhlal.
>>>
>>> Signed-off-by: Glauber Costa<glommer at parallels.com>
>>> CC: Christoph Lameter<cl at linux.com>
>>> CC: Pekka Enberg<penberg at cs.helsinki.fi>
>>> CC: Michal Hocko<mhocko at suse.cz>
>>> CC: Kamezawa Hiroyuki<kamezawa.hiroyu at jp.fujitsu.com>
>>> CC: Johannes Weiner<hannes at cmpxchg.org>
>>> CC: Suleiman Souhlal<suleiman at google.com>
>>
>>
>> Seems reasonable.
>> Reviewed-by: KAMEZAWA Hiroyuki<kamezawa.hiroyu at jp.fujitsu.com>
>>
>> Hmm...but can't we find the 'context' in automatic way ?
>>
> 
> Not that I can think of. Well, actually, not without adding some tests
> to the allocation path I'd rather not (like testing for the return
> address and then doing a table lookup, etc)
> 
> An option would be to store it in the task_struct. So we would allocate
> as following:
> 
> memcg_skip_account_start(p);
> do_a_bunch_of_allocations();
> memcg_skip_account_stop(p);
> 
> The problem with that, is that it is quite easy to abuse.
> but if we don't export that to modules, it would be acceptable.
> 
> Question is, given the fact that the number of kmalloc_no_account() is
> expected to be really small, is it worth it?
> 

ok, but.... There was an idea __GFP_NOACCOUNT, which is better ?
Are you afraid that__GFP_NOACCOUNT can be spread too much rather than kmalloc_no_account() ?


Thanks,
-Kame