Re: [PATCH] memory.min_usage (seqlock for res_counter)

[Pavel Emelyanov <xemul-GEFAQzZX7r8dnm+yROfE0A@public.gmane.org>]

KAMEZAWA Hiroyuki wrote:
> On Wed, 05 Dec 2007 12:12:22 +0300
> Pavel Emelyanov <xemul-GEFAQzZX7r8dnm+yROfE0A@public.gmane.org> wrote:
> 
>> Sorry, let me explain it in other words.
>>
>> I think, that protection in reader, that guarantees that it
>> will see the valid result, is not very important - even if
>> we compare usage and limit not atomically nothing serious
>> will happen (in this particular case)
>>
> Maybe there is no serious situation (now).
> But programmers don't assume that the function may not return trustable result.
> And I think it shouldn be trustable AMAP.

Well... OK. Among other possible ways to achieve this goal
seqlocks is the most preferable one from my POV.

Thanks :)

> I'd like to use seq_lock or res_counter_state, here.
> 
> BTW, I'm wondering I should hold off my patches until 2.6.25-rc series if they
> make things complex.

Actually, Andrew wrote that he will pay little attention to
new functionality till 2.6.24 release, so I think that serious
patches should really be held off.

That's why I don't send the kmem controller yet :(

> Thanks,
> -Kame

Thanks,
Pavel